ACTAUNIVERSITATIS

UPSALIENSISUPPSALA

2013

Digital Comprehensive Summaries of Uppsala Dissertationsfrom the Faculty of Science and Technology 1065

Multicomponent Clusters/Nanoparticles

An Investigation of Electronic and GeometricProperties by Photoelectron Spectroscopy

CHAOFAN ZHANG

ISSN 1651-6214ISBN 978-91-554-8730-0urn:nbn:se:uu:diva-205651

Dissertation presented at Uppsala University to be publicly examined in Room 80101,Lägerhyddsvägen 1, Uppsala, Thursday, October 3, 2013 at 10:00 for the degree of Doctor ofPhilosophy. The examination will be conducted in English.

AbstractZhang, C. 2013. Multicomponent Clusters/Nanoparticles: An Investigation of Electronicand Geometric Properties by Photoelectron Spectroscopy. Acta Universitatis Upsaliensis. Digital Comprehensive Summaries of Uppsala Dissertations from the Faculty of Science andTechnology 1065. 94 pp. Uppsala. ISBN 978-91-554-8730-0.

Clusters/nanoparticles are aggregates of a “small” number of building blocks, atoms ormolecules, ranging from two up to millions of atoms. Two main groups of clusters have beenstudied using photoelectron spectroscopy based on synchrotron radiation. They are dry/wetalkali-halide clusters, including pure water clusters, and metal-based nanoparticles.

For the dry alkali halide clusters, analysis of the data and theoretical modeling has allowedus insights into the local electronic properties at nanoscale: a change of polarizability of ions inthe alkali-halide clusters due to the varying environment has been suggested. The study of thewet salt clusters shows that the alkali-halides are already solvated at the nanoscale reached bywater clusters doped with salt vapor.

The photoelectron angular distribution of water cluster shows lower anisotropy parametersas compared to the separate monomers. A model based on intracluster scattering has been builtto partly explain the reduction.

In the last part of the thesis, metal-based multi-component nanoparticles have been producedby self-assembly processes using reactive magnetron sputtering. Depending on the specificmetal element, oxidation processes have been applied before or after the aggregation. Clearlyradial distributions such as core-shell and “sandwich-like” structures have unambiguouslydetermined by photoelectron spectroscopy.

Keywords: Clusters, Nanoparticles, Alloy, Atmospheric chemistry, Alkali halide, Transitionmetals, X-ray Photoelectron spectroscopy, Polarizability, Core-shell, Sandwich structure,MAX-lab, BESSY II

Chaofan Zhang, Uppsala University, Department of Physics and Astronomy, Box 516, SE-75120 Uppsala, Sweden.

© Chaofan Zhang 2013

ISSN 1651-6214ISBN 978-91-554-8730-0urn:nbn:se:uu:diva-205651 (http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-205651)

To my dear and loving, Yadong

5

Preface

This doctoral thesis presents a number of photoelectron spectroscopic studies on free multi-component clusters/nanoparticles created in-situ, in a beam, by in-house built equipment. Several novel properties appearing at nanoscale have been observed and studied, and are presented in this thesis. All experi-ments have been performed between 2009 and 2013 at the synchrotron radia-tion facilities MAX-lab, Lund Sweden and BESSY II, Berlin, Germany, during my PhD studies in the Division of Molecular and Condensed Matter Physics at Department of Physics and Astronomy, Uppsala University. My PhD work started with the ionic-bonding alkali-halide clusters, first “dry” and then “wet” ones. Then followed a study of the photoelectron angular distribution (PAD) in water clusters. After that the PAD method has been used to study the geometric structure of core-shell lead/lead-oxide nanoparti-cles. At last, the closer to application in optical devices, Yb-based nanoparti-cles of different geometries, from core-shell to “sandwich like” structures have been created and studied. To conclude here, several kinds of nanoscale particles different in geometry and possessing novel properties have been studied. I believe that our work performed in the past four years belonged to fundamental physics, and this work was a very exciting journey for me, and I hope that reading about it will be such a journey for you too. In the follow-ing pages I have tried my best to summarize the beautiful moments in this period. Thank you for reading!

Comments on my own participation Without cooperation, none of the work presented in the thesis could have been accomplished. Concerning my own contributions I was involved in performing all the experiments, model calculations and the following discus-sions. For the papers, except paper III, of which I am the first author, I have been the main responsible for the data analysis and the preparation of the manuscript. I was involved in the measurements and analysis of the data for paper III.

7

List of Papers

This thesis is based on the following papers, which are referred to in the text by their Roman numerals.

I Chaofan Zhang, Tomas Andersson, Svante Svensson, Olle

Björneholm, Marko Huttula, Mikko-Heikki Mikkelä, Maxim Tchaplyguine, and Gunnar Öhrwall. (2011) Ionic bonding in free nanoscale NaCl clusters as seen by photoelectron spectros-copy. The Journal of Chemical Physics 134, 124507.

II Chaofan Zhang, Tomas Andersson, Svante Svensson, Olle Björneholm, Marko Huttula, Mikko-Heikki Mikkelä, Dmitri Anin, Maxim Tchaplyguine, and Gunnar Öhrwall. (2012) Holding on to Electrons in Alkali-Halide Clusters: Decreasing Polarizability with Increasing Coordination. The Journal of Physical Chemistry. A 116, 12104–11.

III L. Partanen, M.-H. Mikkelä, M. Huttula, M.Tchaplyguine, C. Zhang, T. Andersson, and O. Björneholm. (2013) Solvation at nanoscale: Alkali-halides in water clusters. The Journal of Chemical Physics 138, 044301.

IV Chaofan Zhang , Tomas Andersson , Marko Foerstel , Tiberiu Arion , Melanie Mucke , Maxim Tchaplyguine , Olle Björne-holm , Uwe Hergenhahn. (2013) The photoelectron angular dis-tribution of water clusters. The Journal of Chemical Physics 138, 234306.

V Chaofan Zhang, Tomas Andersson, Svante Svensson, Olle Björneholm, Maxim Tchaplyguine. (2013) Core-shell structure in self-assembled lead/lead-oxide nanoclusters revealed by photoelectron spectroscopy. Physical Review B 87, 035402.

8

VI Chaofan Zhang, Tomas Andersson, Olle Björneholm, Xiaojun Xu, Maxim Tchaplyguine, and Zejin Liu. (2013) Variable Ra-dial Structure of Free Nanoparticles by Oxidation Before or Af-ter Aggregation: YbO Shell/Yb Core vs. Yb Shell/YbO Core with Divalent Instead of Trivalent Oxide. The Journal of Physical Chemistry. C 117, 14390 (2013).

VII Chaofan Zhang, Tomas Andersson, Olle Björneholm, Xiaojun Xu, Maxim Tchaplyguine, and Zejin Liu. Alloying and oxida-tion of free core-shell Al@Yb nanoparticles –an “on-the-fly” study (in manuscript)

Reprints were made with permission from the respective publishers. The following is a list of publications to which I have contributed. They will not be covered in this thesis.

1. T. Andersson, C. Zhang, A. Rosso, I. Bradeanu, S. Legendre, S. E. Canton, M. Tchaplyguine, G. Öhrwall, S. L. Sorensen, S. Svensson, N. Mårtensson, and O. Björneholm. (2011) Plasmon single- and multi-quantum excitation in free metal clusters as seen by photoelectron spectroscopy. The Journal of Chemical Physics 134, 094511.

2. Tomas Andersson, Chaofan Zhang, Maxim Tchaplyguine, Svante Svensson, Nils Mårtensson, and Olle Björneholm. (2012) The electronic structure of free aluminum clusters: metallicity and plasmons. The Journal of Chemical Physics 136, 204504.

3. M. -H. Mikkelä, M. Tchaplyguine, K. Jänkälä, T. Andersson, C. Zhang, O. Björneholm, and M. Huttula. (2011) Size-dependent study of Rb and K clusters using core and valence level photoelectron spectroscopy. The European Physical Journal D 64, 347–352.

4. M. Tchaplyguine, T. Andersson, Ch. Zhang, and O. Björneholm. (2013) Core-shell structure disclosed in self-assembled Cu-Ag nanoalloy particles. The Journal of Chemical Physics 138, 104303.

9

5. S. Urpelainen, M. Tchaplyguine, M.-H. Mikkelä, K. Kooser, T. Andersson, C. Zhang, E. Kukk, O. Björneholm, and M. Huttula. (2013) Size evolution of electronic properties in free antimony nanoclusters. Physical Review B 87, 035411.

11

Contents

Preface ............................................................................................................ 5 Comments on my own participation .......................................................... 5

Abbreviations ................................................................................................ 13

1 Introduction ................................................................................................ 15 1.1 Gas-phase clusters/NPs: background ................................................. 15 1.2 Alkali-halide clusters – “dry” and “wet”: particles of environmental relevance................................................................................................... 16 1.3 Metal/metal oxide mixed-composition NPs: Building blocks for nanoelectronics and nanocatalysis ............................................................ 18 1.4 Thesis Outline .................................................................................... 20

2 Photoelectron spectroscopy and related methods ...................................... 23 2.1 Electronic structure of clusters ........................................................... 23 2.2 Photoelectron spectroscopy ................................................................ 24

2.2.1 Electron binding energy spectra and chemical shift ................... 26 2.2.2 Polarization and metallic screening effects ................................ 29 2.2.3 Absolute intensity and relative intensity ..................................... 30

2.3 Electrostatic model for ionic compounds ........................................... 31 2.4 Photoelectron angular distribution (PAD) .......................................... 34

2.4.1 PAD of isolated atoms and molecules ........................................ 34 2.4.2 PAD of clusters ........................................................................... 35

3. Experimental facilities and equipment ...................................................... 37 3.1 Soft X-rays --- synchrotron radiation ................................................. 37 3.2 Experiments at beamline I 411, MAX-lab ......................................... 38 3.3 Experiments at beamline UE 112 / PGM 1, BESSY II ...................... 41

4 Cluster/NPs production .............................................................................. 43 4.1 Adiabatic expansion --- inert-gas cluster source ................................ 44 4.2 Adiabatic expansion --- water-cluster source ..................................... 45 4.3 The modified pick-up source .............................................................. 46 4.4 Gas aggregation --- magnetron sputtering-based source .................... 47 4.5 Reactive sputtering and doping methods of nanoparticle oxidation ... 49

12

5 Results ........................................................................................................ 51 5.1 From dry to wet --- studies of alkali halide clusters ........................... 51

5.1.1 Ionic bonds in the initial and final states .................................... 52 5.1.2 Halogen polarizability at nanoscale ............................................ 57 5.1.3 Wet salt clusters --- solvation at nanoscale ................................. 60

5.2 Reduced anisotropy in the photoelectron distribution from clusters and NPs .................................................................................................... 62

5.2.1 Decrease of anisotropic parameters in water clusters ................. 62 5.2.2 Elastic scattering effects for metal bulk atoms ........................... 66

5.3 Core-shell structure of metal/metal oxide NPs ................................... 67 5.3.1 Disclosing core-shell structure of multi-component NPs using photoelectron angular distribution ....................................................... 68 5.3.2 A@B and B@A core-shell NPs formed of Yb and YbO............ 70 5.3.3 Core-shell bi-metallic nanoalloy YbAl ....................................... 72 5.3.4 Oxidation of nanoalloy YbAl particles --- the sandwich NPs .... 74

6. Summary & Outlook ................................................................................. 77

Sammanfattning på svenska .......................................................................... 81 Kluster / Nanopartiklar ............................................................................. 81 Elektronisk struktur och fotoelektronspektroskopi .................................. 82 Resultat vi fick ......................................................................................... 83

Bibliography ................................................................................................. 85

Acknowledgments......................................................................................... 91

13

Abbreviations

NPs Nanoparticles PAD Photoelectron angular distribution BE Binding energy

15

1 Introduction

Particles of nanoscale were discovered by people only very recently in mod-ern history. In such particles, novel properties emerge due to the unique ef-fects –mainly quantum effects that only appear at this scale of dimensions [1]. A special feature of the studies in this thesis is that they have been carried out on unsupported nanoparticles (NPs) –in vacuum environment. The prop-erties of the NPs were studied indirectly by detecting the photoelectrons after the ionization by soft-x-rays from a synchrotron facility. In the Introduction, I would like to present some background information on the different kinds of NPs investigated in this thesis.

1.1 Gas-phase clusters/NPs: background At the nanoscale, the properties of matter can be different, for example, at sub-micrometer sizes, colloidal gold can be red (smaller than 100 nm) or blue/purple (for larger particles), e.g. the colour is dependent of particle size[2]. The changing colour is just one example of the special properties of nanoscale objects which we will discuss in this thesis. Clusters/NPs are ag-gregates of a “small” number of building blocks, atoms and molecules, rang-ing from two up to millions of atoms [3]. As the size of NPs is between the molecule and solid, the properties of NPs can be completely different from the corresponding molecule or solid. Moreover, the chemical and physical properties of NPs may vary significantly with the size [4,5], as well as the spatial distribution of constituent atoms created in the cluster formation process [6]. NPs can be seen as an intermediate phase of matter, and to a great extent because of this they constitute a very interesting topic to study.

Over the history of the NP-science numerous kinds of NPs have been cre-ated and studied under laboratory conditions, such as, for example, inert-gas NPs [7–9], NPs out of molecules, metal NPs [10], alloy NPs[11], NPs contain-ing metal with organic molecules [12] etc. But do NPs exist only in a labora-tory? Of course not, actually NPs exist everywhere. For example, in the at-mosphere [13], there is a lot of natural NPs. Even more NPs have left labora-tories and became building blocks for advanced devices that revolutionarily have changed the performance of the technology surrounding us, and in the end, may have changed the lifestyle of people. NPs have been applied in various fields, such as solar cells [14], catalysis [15], optical devices [16],

16

medical engineering [17], etc. More potential applications are promised to come out from the study of various special NPs.

In this work, we try to create several different kinds of NPs, which can be mainly divided into three groups: the first one is the ionic-bonded alkali-halide clusters, first as dry aggregates, and then as complexes with water molecules. Another group is the metal alloy and metal oxide NPs. Presented in between these two groups, water clusters represent a class of its own. To follow the traditionally established terms in somewhat different scientific communities, in our thesis, we call the particles of alkali halide and water molecules as clusters, and the metal containing particles as NPs. Actually they are the same type of objects –in spite of different names.

1.2 Alkali-halide clusters – “dry” and “wet”: particles of environmental relevance In the atmosphere close to the ocean surface, the water spray droplets con-tain alkali-halide ions. While in the sea water the alkali-halides exist in the form of free ions [18], in the process of evaporation, the water and alkali-halides separate to a great extent, resulting in drier alkali-halide particles. The schematics of the formation of dry and wet alkali-halide clusters is shown in figure1.1. Both types play an important role in the formation of the climate of the Earth and for photo-chemical processes in the atmosphere. In particular, these clusters serve as seeds for cloud condensation, as scattering centers for VUV, UV, visual and infrared radiation, as constituent parts in the larger particles containing sulfates, hydrocarbons, ammonia, etc. Such an active role of the alkali-halide clusters is defined by their reactivity deter-mined in its turn by their composition and electronic properties. While the alkali-halide water solution and crystalline solid have been extensively stud-ied, see, for example, [19,20] and references therein, the other two phases (dry and wet clusters) existing in the atmosphere are much less well known. In this thesis, I will present a study of these two phases, the dry and wet alkali-halide clusters performed in our group. There is one more reason why alkali-halide clusters attract the interest of researchers: they are the simplest ionic compounds with still countable number of constituent atoms.

17

Figure 1.1 Schematic illustration of wet and dry alkali-halide clusters formed in the atmosphere above the sea

The clue to understand the physical and chemical properties of such nano-scale particles is to study their electronic and geometric structure. Photoelec-tron spectroscopy is a powerful tool to directly investigate the electronic structure of the sample [21]. The common way to study solid matter by pho-toelectron spectroscopy is to ionize the sample mounted on a grounded sub-strate. However, in our case, the alkali-halide clusters are insulators. Serious problems arise for such supported sample: it charges as the electrons leave it, severely reducing the interpretability of the spectra. Thus, one advantage of the present work is that the alkali-halide nanocrystals have been created in a beam – in a form of continuously renewed free clusters, generated by a dedi-cated cluster source. At our x-ray intensity levels each cluster is photoion-ized only once in the experimental region and leaves this region very fast. Therefore the charging effects are eliminated.

The geometric structure of dry alkali-halides has earlier been studied by various other experimental techniques combined with calculations, see, for example [22–24] and references therein. However, there are still many aspects to be investigated. In addition the geometry of wet alkali-halide clusters is to the best of our knowledge still far from being completely clear. Are the molecules dissociated into separate ions at nanoscale as in macroscopic solu-tions? If so, is there any segregation in the ion positions, for example, sur-face enrichment for this or the other type of ions in wet alkali-halide NPs?

18

These are the questions which electron spectroscopy can address. In the works included in this thesis, both dry and wet clusters have been produced under controlled laboratory conditions, and photoelectron spectroscopy has been employed to measure the main observable of such an experiment; the binding energies of the core- and valence-level electrons. This information allows us to draw conclusions not only on the electronic structure, but also on the mutual positions of the constituent atoms. An electrostatic model has been constructed to help interpreting the electronic and deriving the geomet-ric structure reflected in the experiment. Analysis of the data and theoretical modeling has allowed us to obtain a deeper insight into the local electronic properties at nanoscale: a change of polarizability of ions in the alkali-halide clusters due to the varying environment has been suggested as one important mechanism. Finally, the comparison of the results on the dry clusters with those on the wet clusters has allowed us to shed some light on the structure of wet clusters for several alkali-halides in this work.

In order to produce wet clusters, the water clusters were produced first by the adiabatic expansion technique [25,26]. And in principle, water clusters themselves are very specific matter to study. Researchers try to understand the interaction of water molecules in liquid water by studying the clusters with limited number of molecules [13,27,28]. Water molecules are kept together in water clusters by hydrogen bonding [29]. Many aspects of this bonding are still not fully understood. Therefore, in our work on water clusters included in this thesis, we have also tried to unveil some mysteries of water studying angular distribution (PAD) of photoelectrons ejected from the clusters as the result of x-ray irradiation. Earlier a simple elastic scattering model has been shown to be enough to explain the different (relative to separate monomers) PAD behaviour in some other types of clusters. However, it is not enough for water cluster, as our results show. We have not been able to comprehen-sively explain all the details of the PAD behavior in water clusters. However, to my opinion, it does not diminish the value of this kind of study and the results obtained.

1.3 Metal/metal oxide mixed-composition NPs: Building blocks for nanoelectronics and nanocatalysis The second group of NPs which are discussed in the thesis are metal-based. In most of the works, this kind of NPs is formed on substrates via atomic vapour deposition [30], aggregation in solutions [31] or, as it is done in our group, via condensation in the cooled gas-mixture containing metal vapour [11,32,33]. NPs of this type are discussed for various applications and are in many cases already used. Compared to the macroscopic materials, noticeably higher performance could be reached for NPs, for example, in getting

19

stronger luminescence by substituting traditional material to quantum dots [34]. Similarly to the alkali halide clusters, NPs containing metals can also exist in atmosphere however mainly as ingredients of toxic pollution. It has been reported that the inhalation of high concentration of NPs is harmful to the health of people [35]. In the works included in this thesis, we have tried to fabricate free metal-based NPs, in a particle beam. Therefore, this study on the one hand meets the demand for novel materials from technology, and on the other hand gives a better understanding of such NPs formation and struc-ture what is also beneficial from the environmental perspective. Moreover, the experiment on the gas-phase NPs makes it possible to study a pure mate-rial in vacuum -without the otherwise always present influence of surround-ings.

Figure 1. 2. Schematic illustration of NPs with different radial distributions: the core-shell structure of metal/metal-oxide, segregated nanoalloy, and “sandwich-like” structures.

The production of NPs discussed in the thesis has been realized via self-assembly. The geometric structure is an important property of NPs. The most common contemporary way to obtain information on it is to image them using microscopy. In our work such information has been obtained in an indirect way elaborating the signal from the emitted photoelectrons. The basic geometric structure of NPs can be cubic (like for ionic clusters) [36], tetrahedral [37], quasi spherical [38], etc. In this study, we are not able to strictly determine the geometry of the NPs. As suggested by different stud-ies, the shape of NPs we produced can be often roughly approximated by a sphere [38]. Only knowing the shape, however, is not enough. Perhaps even more important is the distribution of the elements in multicomponent NPs,

20

which is also the main issue in the studies presented in this thesis. The distri-bution can be uniform, with some segregation, core-shell, sandwich-like, or with other internal element distribution. The mixed composition NPs studied in this thesis are schematically shown in figure 1.2. In this thesis, the metal based NPs were produced by gas-aggregation utilizing magnetron sputtering (see for details below). The NPs can be expected to have a radially symmet-ric element distribution due to the symmetric creation process. For core-shell NPs, at first, several different metal/metal-oxide mixed composition NPs have been studied, starting with Pb/Pb-oxide. In the Yb/YbO NPs two mutu-ally reversely ordered types of NPs have been also realized using two differ-ent methods of oxidation. The other core-shell NPs studied in the thesis are the nanoalloy particles of Yb and Al. Clear separation of two metals in these latter NPs has been observed. The last type of NPs studied in this thesis is the NPs in the intermediate stage of oxidation with the oxidized surface of Yb, metallic Yb interface, and Al metallic core –the “sandwich-like” NPs.

X-ray photoelectron spectroscopy (XPS) is a powerful tool capable of re-solving responses from different sites in NPs [39]. The same atoms but be-longing to different sites, like surface and bulk of the NPs, can be distin-guished by XPS due to different energy positions - and relative intensities in the spectra. However, for some substances, including some metals, the sur-face and bulk binding energies are very similar, like in lead. In this case we have used the photoelectron angular distribution (PAD) to deduce the core-shell structure of Pb/Pb-oxide mixed-composition NPs.

In the works included in the thesis the oxidation of metal NPs has been performed before and after the aggregation. The vapour of metals has been created by the sputtering method. In the case with oxidation before the ag-gregation the so-called reactive sputtering has been implemented, and in the case with the oxidation after the aggregation certain type of oxygen doping has been utilized. By using these two methods, it has been possible to fabri-cate various kinds of NPs. The photoelectron spectroscopy applied to the unsupported NPs has allowed us to determine their structure with a high degree of probability.

1.4 Thesis Outline The thesis is divided into two parts, I and II. In addition to this introductory-chapter 1, part I will continue with:

• in chapter 2, a description of the probing experimental technique (photoelectron spectroscopy) used for the studies, and some related methods involved in the thesis.

• in chapter 3, a description of experimental facilities and equipment.

21

• In chapter 4, a description of the process of the cluster production. The alkali-halide clusters have been produced by the so-called pick-up cluster source in two steps. In the first step two different setups have been employed for producing clusters: A source for inert-gas clusters, and a water-cluster source. Both sources are based on the adiabatic expansion process. The metal mixed-composition NPs were produced by magnetron sputtering-based gas-aggregation source.

• in chapter 5, the presentation of experimental results obtained by photoelectron spectroscopy in combination with the electrostatic model disclosing the electronic and geometric structure of alkali-halide clusters ( Papers I and II). We discuss how the polarizability of ions in the clusters varies with the site and size (Paper II), and fi-nally we summarize the experimental results on the solvated alkali halide clusters (Paper III). The second part of chapter 5 is about the study of water clusters and Pb/Pb-oxide NPs using PAD. The third part describes the study of core-shell NPs of metal and metal oxides.

• in chapter 6, a summary and an outlook.

Part II, contains papers I-VII, published and in manuscript.

23

2 Photoelectron spectroscopy and related methods

2.1 Electronic structure of clusters Clusters can be seen as an intermediate type of matter with properties in between those of separated atoms/molecules and the macroscopic bulk. The electronic structure of clusters is correspondingly different from that of both traditional matter types. For a separate atom, the electronic state can be de-scribed by a wave function, and such a wave function is determined by series of quantum numbers, such as the principle quantum number, the angular momentum quantum number, the magnetic quantum number, and the spin quantum number. In principle, for each set of quantum numbers, there is a separate level, occupied or unoccupied. Figure 2.1, schematically shows the energy scheme for a separate atom, for simplicity with only one occupied and one unoccupied level. For a simplest separate atom with only one elec-tron, the energy between the highest occupied level and the vacuum level is the ionization potential, which defines the lowest energy required to remove an electron from the atom. In terms of photoelectron spectroscopy this is also the binding energy of such electron. For an electron in a system with more than two electrons, this energy is not the same as the orbital energy due to the relaxation of remaining electrons upon ionization.

When the number of atoms in a system becomes equal to two (a molecule is built), in a simplified picture one can think that the states for two atoms, both occupied and unoccupied are split into two -due to the interaction be-tween the atoms. However, the states are still well discrete. For more atoms in a system, all the discrete states become denser and one can talk about different bands of discrete states, unoccupied and occupied. In the macro-scopic bulk material, also in metals, the different bands become continuous. A small cluster containing just several atoms resembles in its energy struc-ture a complex molecule. With the number of atoms increase the electronic structure of clusters changes from molecular-like to bulk-like, from discrete to band-like structure, when the number of the constituents becomes suffi-ciently large. The band of occupied states corresponds then to the valence band in macroscopic solids, and the band of unoccupied states – to the con-duction band. When clusters are not metallic there is a gap between the two bands. In metal clusters the lowest energy necessary to remove the electron

24

from the system can be seen as the cluster work-function-analogous to that of the “infinite” solid metal. The transformation of the energy structure dis-cussed above is schematically presented in figure 2.1.

If the gap between the conduction band and valence band is absent, then the cluster is conductor, like metals. If the gap is very large, then it is an insulator. Smaller metal clusters can be semiconductors, for which the gap is not yet “closed”, and moreover can be changed by varying the number of atoms in the cluster.

Figure 2.1. Schematic illustration of the electronic structure transformation from separate atoms to the macroscopic bulk metallic materials.

2.2 Photoelectron spectroscopy The experimental techniques in this study are all based on photoelectron spectroscopy. As not once mentioned above, photoelectron spectroscopy is a powerful tool to investigate the physical and chemical properties of various samples. The principle of photoelectron spectroscopy is based on photoioni-zation of an atom/molecule/solid/etc. As the result of the photon absorption an electron is emitted with a definite kinetic energy. By measuring the ki-netic energy of the electron, the binding energy of the electron can be deter-mined according to Einstein’s photoelectric effect law:

( ) binifkin EhEEhE −=−−= νν

Here h ν is the photon energy, fE and iE are the final and initial state

total energies of the system, and bin f iE E F= − is the electron binding en-

ergy. The kinetic energy can be measured by an electron energy analyzer, as

described above. Usually several ionized final states exist in the probed

25

binding energy region. Thus, the electrons are emitted in ionization-

transitions between the ground and these various ionized states. These elec-

trons have different kinetic energies which together constitute a photoelec-

tron spectrum.

Figure 2.2 Schematic illustration of the valence level (left) and the core-level photo-electron spectroscopy (right).

If the h ν is not high enough, the incoming photons may promote a bound electron from a core or valence state to a higher unoccupied state. This proc-ess is regarded as photoexcitation. In both direct photoionization and photo-excitation, the fast electronic excitation decay could follow as localized (in dielectric) or delocalized (in metals) Auger decay or cluster-specific delocal-ized Interatomic Coulomb Decay (ICD) [8,40].

The synchrotron radiation beam crosses the cluster beam (which produc-tion is described in detail below) at an angle of 90°. The electron energy analyzer –a device catching the emitted electrons and determining their ki-netic energy - can be placed at different angles relative to the polarization plane of the x-ray radiation. (The radiation is linearly polarized in all the experiments described) The photoelectron spectra of the present work have been in most cases recorded with the analyzer fixed at so-called “magic an-gle” and at 90°. In the experiments on water clusters we could also measure the photoelectrons emitted at other relative angles by using a special type of undulator [41] with the rotatable polarization plane of the radiation.

After the spectra have been recorded by the electron energy analyzer, the different features in the spectra can be used to interpret the physics behind the transitions, to obtain information about the initial and the final states of

26

the system under investigation. The energy positions and the absolute and relative intensities of the features carry this information.

2.2.1 Electron binding energy spectra and chemical shift If we study an isolated atom, the photoelectron spectrum reveals the elec-tronic structure of the initial and the final states. In a typical case the spectral responses are sharp peaks with well-defined Voigt shapes, which is a convo-lution of Lorentzian and Gaussian spectral profiles. The Lorentzian shape is due to the finite life-time of the ionized state of an atom. The width of the Gaussian peak is determined by the resolution of the light source and the electron energy analyzer, as well as by the existence of various local envi-ronments and the thermal motion of the atom. Therefore, the Gaussian width, EΔ , could be calculated according to the formula:

2 2 2m sp thE E E EΔ = Δ + Δ + Δ

Here mEΔ is the width of incident radiation, which is defined mainly by

the monochromator slit, spEΔ is the width due to the limited spectrometer resolution and thEΔ is the width due to thermal broadening.

In the present study of the metal NPs, the spectra are closer to the metal macroscopic solid. The photoionization of the metal NPs can induce elec-tron-hole pairs near the Fermi-level, which may reduce a part of the photo-electron kinetic energy. Therefore, compared to the symmetric shapes in the atomic spectra, the spectral “lines” of the metal NPs are asymmetric with a tail towards lower kinetic energies. The asymmetry is defined by the electron density of states at the Fermi level and degree of the final state screening after the photoionization. This kind of spectral lines can be described by an asymmetric line shape given by the Doniach-Sunjic formula:

( )[ ] 2

122

0

01

0

)(tan)1(2

cos

),,,( α

γ

γαπα

γα −

−

+−

−−+=

EE

EE

EEDS

The energy position of the peak directly gives the binding energy of the electrons. One usually distinguishes between valence and inner or core elec-trons. The valence levels can often in their turn be divided into the inner- and outer- valence ones. If the outermost valence electronic state is ionized, the atom can exist in such a state for a theoretically infinite amount of time τ (provided that atoms are completely isolated). Ionization from all other orbi-tals, inner-valence and core ones, gives ionic states with a finite life-time,

27

since there is always a probability of either electronic or radiative decay. According to Heisenberg’s uncertainty principle the life-time determined Lorentzian width is ~1/τ. Therefore, if the lifetime is very long the Lorenzian linewidth is negligible in comparison to the Gaussian experimental broaden-ing. As a result the Voigt profile reduces to a Gaussian. If instead a core level is ionized, the vacancy created in the core shell is filled very fast, typi-cally on the low femtoseconds time scale, by one of the valence electrons. Thus the life-time of the ionized state is shorter, and the corresponding width is larger than that of the valence levels.

Already for the simplest di-atomic molecules the photoelectron spectra of both valence and core levels are much more complex than in atoms. Vibra-tional and rotational states manifest themselves in both types (valence and core) of spectra for the molecules, and even for the core-levels there can be complications such as partial delocalization of the core holes and molecular field splitting. For the valence levels, typically, the more complex the mole-cule is the broader and less resolved the corresponding spectrum becomes. For the core levels it is somewhat simpler, since core orbitals are atomic-like. The difference in energy between the core-levels in different chemical sur-roundings is due to the different distributions of the valence electrons in both the initial and final states. The binding energy of the core-level, which is determined as the difference Ef - Ei, does not only depend on the form of the core electron orbitals, but on all the orbitals of the system under investiga-tion. The change in the binding energy from the atomic energy to that in a molecule or solid containing the same atom is often called a “chemical shift” (see figure 2.3). In some case, the chemical shift is also used to denote the binding energy difference between the same orbitals in different molecules or sites.

When an even more complex system is formed out of many separate at-oms, like a solid, the valence electron orbitals are replaced by wide energy bands, in which the electron energy is dispersing with the momentum. In this work, we study cases in between the molecules and solids with a finite amount of atoms/molecules bound in a cluster. Very often a large cluster can be seen as a nanoscale piece of the corresponding solid. However, the sys-tem may in a certain sense be simpler than the solid, what in some cases makes it possible to deduce more information from a photoelectron spectros-copy study on clusters than from a study on a corresponding solid. Espe-cially informative can be a study of free, non-supported clusters. In such a case for the core-levels, the chemical shift in the binding energy for specific atoms in a cluster (nanoscale solid) can be recorded without being disturbed or obscured by a macroscopic substrate.

28

Figure 2.3 Schematic drawing illustrating the chemical shift for metal electron bind-ing energies between the molecule and ionically bound cluster.

There could be several mechanisms behind a chemical shift in compounds. For the clusters of ionic molecules, the main effect comes from the Coulomb interaction of the atom/ion under investigation with the ions of the neighbour molecules in the cluster, as well as from the polarization interaction. Relative to the corresponding molecules, both the initial and the final states are changed in such clusters. Interactions of the following types have been con-sidered in the present work:

1. In alkali-halide clusters, both Coulomb and polarization interaction takes place between the probed ion/atom and its neighbour ions.

2. In the case of wet cluster production the solvated alkali-halide mole-cules interact with the water molecules in the complex cluster. Ions of the salt induce polarization on the neighbouring water molecules cre-ating a solvation shell and changing the initial and the final state en-ergies of the salt ion under investigation.

29

All the interactions above will shift the binding energy of photoelectrons emitted from alkali-halide atoms, relative to the case of a separate salt mole-cule. The Coulomb interaction in salt crystals is since long a subject of un-dergraduate chemistry textbooks. The polarization screening interaction in the case of a finite number of atoms still deserves a more detailed treatment and will be described below.

For a metal cluster/nanoparticle the core-level binding energy values can deviate from the value for the corresponding macroscopic sample due to the influence of the various initial and final state effects. In the initial state the binding energy is influenced by the metallic bonding between the atoms. For the metallic NPs, the complete screening by the delocalized valence elec-trons appears as the final effect after photoionization. When the metal is oxidized, the binding energy changes compared to the pure metal, which is due to the appearance of the surrounding oxygen atoms. This influences both the initial and the final state energies. The changes can be different depend-ing on the metal coordination to oxygen, in other words –depending on the interior structure of the NPs. For the lead oxide NPs in this work, the binding energy of the oxidized Pb does not change at different oxidation conditions, but, in general, different oxides usually manifest themselves at different binding energies. Moreover, it is possible to study the NP composition: the distribution of the oxide and metal - by comparing their relative intensities in the photoelectron spectra.

2.2.2 Polarization and metallic screening effects For the alkali halides, in the initial state of an alkali-halide molecule the metal and halogen are bound together by ionic bonding: the metal atom has given away one electron to the halogen atom which thus is negatively charged. Additionally, the field of the ions induces dipoles on the counter-ion by distorting the electron charge distribution on them. In the process of core-level photoionization the metal atom loses one more electron. When the halogen atom is core-ionized it also loses an electron and is neutralized. So for the case of metal core-ionization, the final-state charge distribution of the neighbours is stronger polarized by the newly formed doubly charged metal atom, whereas the polarizing field existing in the initial state disap-pears in the final state for the case of the halogen-atom core-ionization. In both initial and final states the polarization-interaction energy changes the binding energy of the electrons relative to its value in free ions.

In a cluster of molecules the number of neighbours is considerably larger than in a molecule: so the role of the polarization interaction increases. The total decrease in the metal binding energy occurs to be about 3 times larger in a small cluster relative to that in a dimer. For the halogen atom in clusters we have observed a spectral response shifting to higher binding energy rela-tive to that in a molecule, due to both the Coulomb and polarization energy

30

being absent in the final state. More details will be given in the section de-scribing the ionic model for cluster ionization. The detailed analysis of the experimental observation of these phenomena will be given in the “Results” section.

In the case of alkali-halide molecules picked-up by a beam of water clus-ters, our results support the scenario of the alkali-halide molecules being dissolved in water clusters and existing as free ions. The ionic bond is bro-ken and the metal and halogen ions are surrounded by water molecules. In such a case the polar water molecules, especially in the first solvation shell, orient themselves along the electric field of the ions. Both the inherent dipole moment of water molecules and the additional field-induced dipole should contribute to the polarization interaction with the ions, and thus to the screening effect created by the water molecules. Therefore, the electrons in the solvated ions also get different electron binding energies.

For metallic NPs, the outer valence electrons exist as free, delocalized electrons. After the core-ionization by the x-ray photon, a core hole is cre-ated in the atom, followed by rapid interatomic charge rearrangements, i.e. one of the neighbouring free electrons gets localized to the core-hole site. According to the Z+1 approximation method, this results in a core ionized atom with one extra valence electron, i.e. resembling the next element in the periodic table. This process is called the complete, metallic screening of the “core-hole”. Since one of the free electrons has been localized at the core-hole site, there is now a “hole” among the free conduction electrons. Since the NP is metallic, this “hole” will spread out around the surface of the NP. Therefore, there is no polarization screening effect in the bulk metal and larger NPs. However, there is also some special case, when the polarization screening effect has been observed for the metal NPs in the case when the size is small enough for the NP to be non-metallic (e.g. less than 20 atoms in the Pb clusters) [42].

2.2.3 Absolute intensity and relative intensity As briefly mentioned above, certain aspects of the physics behind the spectra can be obtained by analyzing the absolute and relative intensity of the peaks in the spectrum, for example, of those corresponding to different sites. The use of information on relative intensity can be illustrated by the case when bulk and surface responses are separated in photoelectron spectra, what has been known since long for solids and has been recently studied [5,11,43,44] also for free clusters/NPs of different sizes. For example, the structure of bi-component metallic clusters has been investigated by comparing the relative intensities of bulk/surface responses of different constituent atoms [11]. The electrons emitted from the deep bulk may lose their energy due to the inelas-tic scattering on the cluster-constituent atoms, therefore reducing the signal from the bulk. In a mixed cluster, the surface enrichment by one of the ele-

31

ments can be studied by comparing the relative intensity of the bulk and surface responses for the pure and the mixed-composition cases.

In the studies presented in this work, the alkali halide clusters have a cu-boid geometry [23,24,45]. As our study has shown, the spectral responses of the atoms at different sites (corner, face, edge, and bulk) of such a cuboid are significantly different: they are separated in energy and have different rela-tive intensity. This can be explained by the number of neighbours for these sites being different. In the spectra, as the size of clusters changes, the rela-tive intensity of the response from different sites also changes. For larger clusters, the intensity due to the face atoms and bulk atoms increases relative to that due to the edge atoms. Because the responses to a significant extent -overlap, the resulting peak composed of the sum of the contributions from the different sites will shift as a whole. For metal-atom response alkali-halide clusters it shifts towards lower binding energy with the cluster size. More details will be described in Chapter 5- Paper II.

For some metal NPs, like those of Yb in this work, the bulk and surface features are well resolved. And after the oxidation of these NPs, not only the binding energy shifts but the relative intensity of the surface-bulk changes due to the emergence of the new sub-components. By judging from the rela-tive intensity of the surface and bulk responses, the oxidation process details can be clarified. In its turn it provides information allowing us to disclose the component distribution in the mixed-composition NPs.

2.3 Electrostatic model for ionic compounds The alkali-halide monomer MX core-level energies can be closely predicted using the electrostatic approach for bonding between the positive metal ion M+ and the negatively charged halogen atom X- [46]. As mentioned above, in addition to Coulomb interaction, one has to consider the energy due to the polarization of the electron clouds in both the initial and the final states in-duced by the charges. This contribution when taken into account in the cal-culation leads to a good numerical agreement with the experiment.

For a molecule the polarization-energy can be written as ½q2α/r4, where q is the inducing charge separated by a distance r from the counter particle, and α is the polarizability of the counter particle. In the initial state, both positive and negative charges polarize the electron distribution of the neighbour sites. In the final state, in the case of alkali metal ionization, the metal ion becomes doubly charged, which leads to a twofold increase of the Coulomb attraction and consequently to a quadruple increase of the polariza-tion energy. For halogen atom ionization, the charge is zero in the final state, so as the ion becomes neutral, the Coulomb interaction disappears, and no dipole is induced on M+.

32

As mentioned above, the binding energy of an electron equals to the dif-ference between the final state energy and the initial state energy of the sys-tem as a whole. In an M+X-molecule it is:

ΔWmol= Wf,mol - Wi,mol.

The final state and initial state energy, for the metal site as an example, can be given as follows:

Wi,mol(M)= Wat(M+)+WCoulomb, i+Wpolarization, i

Wf,mol(M)= Wat(M2+)+WCoulomb, f +Wpolarization, f

Here Wat(M+) and Wat(M2+) are the isolated-atom energies in the initial and final states, respectively. These values are well known.

As discussed above, the detailed terms of the Coulomb energy and polari-zation energy in the initial state and in final state can be written as:

WCoulomb, i = r

e2

− (4) and WCoulomb, f =r

e22−

Wpolarization, i = 4

2

4

2

22 r

e

r

eMX +−

−−αα

Wpolarization, f = 4

2

4

2

22

4 2

r

e

r

eMX +−

−−αα

Where αX-, αM+ and αM2+ are the polarizabilities for correspondingly M+,

X- ions of the initial state, and M2+ ion of the final state. For an alkali-halide cluster consisting of a certain finite number of atoms

the case becomes more complicated. However, some necessary information has been known since earlier: it has been established with a large degree of probability that the geometric structure of alkali-halide clusters is that of cubes or cuboids [23,47]. Assuming such a structure one can approximate the total energy and the ionization transitions using similar steps as in the mo-lecular case. The Coulomb energy can be calculated as the sum of the Cou-lomb interaction of the probed atom with all the neighbour atoms.

WCoulomb, i = n

iCoulombnW ,,

WCoulomb, f =n

fCoulombnW ,,

For the polarization interaction, we first have to calculate the total electric

field nE

on each of the atoms, and then the value and the direction of its

33

dipole np

is determined as nnE

α , where αn is the n-th ion polarizability

(either of X-, or of M+); then the electrostatic potential Un of each np

dipole

on the chosen atom is calculated as a scalar product of the dipole np

and the

radius-vector nr

drawn from the dipole to the chosen atom: Un =3

/ nnn rrp •

. This allows us to determine the energy lowering due to the interaction of

the charge of the chosen atom q (M+, M2+ , but also for X-) with each dipole

created by the total field on each n-th atom as Wn=½qUn; the total energy

due to the interaction between the charge q and all dipoles is calculated as a

sum of individual energies Wn; There is one more energy-lowering contribution to calculate due to the in-

teraction of the dipole induced on the chosen atom by the field of all other atoms with the charges of all other atoms. Here one computes:

- the vector nE

of the Coulomb field of each n-th ion at the chosen-atom position;

-the total field on the chosen atom as a sum of all nE

; -the dipole appearing on the chosen atom p

=

≠chosenn

nE

α , where α is the

polarizability of the chosen-atom;

-the electrostatic potential Un of the dipole p

imposed at the position of each atom in the cluster as a scalar product

3/ nn rrp • where nr

is now the

radius-vector from the chosen atom to each n-th atom; -the interaction energy of the dipole with each ion qn as Wn=½qnUn;

- the total energy lowering as ≠chosenn

nW .

After we obtain the total Coulomb and polarization energies for the initial and final states, respectively, the core-level binding energy of a specific-atom electron can be estimated using this electrostatic model. The size- and site-dependent properties can also be predicted by the model.

The model calculation for the halogen atoms is similar and simpler due to the disappearance of the charge on the core-ionized halogen in the final state. As the result the contribution of the “charge-charge” interaction is deter-mined only by the initial state. So for the Coulomb energy, the change be-tween the final state and the initial state is:

WCoulomb, f -WCoulomb, i = r

e2

For the polarization energy, the final state has only one term:

34

Wpolarization, f = 4

2

2r

eX −

−α

The total change of the polarization energy is:

Wpolarization,f -Wpolarization, i = 4

2

2r

eM +α

Comparing the ionization of a molecule and a free atom along the lines presented above, one obtains the following binding energy shift for the halo-gen atom in the molecule:

ΔW(X) =r

e2

+4

2

2r

eM +α

Similarly one can calculate the shift from the monomer (molecule) spec-

tral response to the cluster response. We observe the same spectral behaviour in the experiment as we have obtained from the model prediction.

This model has been used for the sodium and potassium halide clusters in paper I and II.

2.4 Photoelectron angular distribution (PAD)

2.4.1 PAD of isolated atoms and molecules For the photoionization of isolated, randomly oriented atoms or molecules with fully linearly polarized light, the angular dependent cross section

),( φθp of emitted photoelectrons can be described as [48]

( ) .)(2cos314

14

),(

−++=

Ω= λθβ

πσσφθ P

d

dp

Here, , σ is the total photoionization cross section, θ is the angle between the horizontal and the photoelectron emission direction (within the dipole plane), λ is the angle of the major axis of the polarization ellipse to the hori-zontal, P is the degree of linear polarization, and β is the anisotropic parame-

35

ter. Experimentally, the β parameter for each specific orbital can be deter-mined from the measured angular dependence of the signal. β is bounded between −1 and 2. A higher β means a larger differential cross section in the direction parallel to the polarization vector of the radiation. In other words, the emission of electrons is more anisotropic. β = 0 corresponds to isotropic emission. As can be shown using the formula above, there is a certain angle, the so-called “magic angle”, of 54.7°at which the differential cross-section becomes independent from the angle.

2.4.2 PAD of clusters A question can be asked whether the PAD in clusters is the same as in the constituent monomers. Earlier experimental studies on the inert-gas clusters have shown that the anisotropy parameter β decreases in clusters [7,48,49]. Clearly, the electronic structure of clusters is much more complicated than that of the constituent monomers. Many known and unknown effects could play a role here for PAD. For example, the PAD of the same atoms but posi-tioned at different sites could be different. The unambiguously established effect influencing PAD is the intracluster scattering, and it can be calculated with a simple model. The intercluster scattering is neglected in this ap-proach.

In the works included in the thesis, we only measured 2-dimensional PAD of water clusters, so the model is not dependent on the azimuthal angle φ. At first, we define the differential scattering cross section for a constituent monomer )(θf , and with the normalization as:

σθθθφπ π

= sin)(2

0 0fdd

Therefore, the amount of electrons scattered into a given direction of ob-servation r̂),( ≡φθ can be written as a convolution of the original angular distribution function )ˆ(rp ′ with the scattering probability f . Since f de-pends on the relative angle )ˆ,ˆ( rr ′∠ we have

( ) ))ˆˆ(arccos()(sin))ˆˆ(arccos()ˆ(ˆ)ˆ( rrfpddrrfrprdrpf ′⋅′′′′=′⋅′′=⊗ θφθθ

Here the relative angle )ˆ,ˆ( rr ′∠ depends on φ ′ , therefore the integration

over φ ′ is not trivial.

The secondary scattering process does not refer to any fixed direction in

space. It therefore cannot lead to a change of the general functional relation-

ship of the PAD, but only to a change of the observed value of the β parame-

36

ter. We call the angular distribution parameter of an electron ensemble after

single scattering for βf. In an experiment, the unscattered part of the photoelectron flux is ob-

served with its original β, while another part, scattered inside the cluster, is observed with βf. To obtain an estimate for the probability of single scatter-ing, a structural model of the clusters is needed. For the fraction of the elec-trons scattered, we have )exp(1 lnq σ−−= , and this can be integrated by the model based on the geometry of the clusters. The anisotropy parameter β of clusters can be calculated as

βββ )1((mod) qq fcl −+=

This model has been used for the study of water clusters in Paper IV.

37

3. Experimental facilities and equipment

All the clusters/NPs in the publications presented in this work have been studied by photoelectron spectroscopy using synchrotron radiation. To im-plement such an approach, we need a light source providing the photon en-ergies capable of ionizing the sample, and an electron kinetic energy ana-lyzer. The experiments have been mainly performed at the Swedish synchro-tron radiation facility MAX-lab and partly at the German synchrotron radia-tion facility BESSY II, where both necessary ingredients (the light source and the energy analyzer) are at the user’s disposal at several beamlines. This chapter is intended to provide a brief background of the experimental tech-niques used in the presented work. Detailed introductions to these techniques can easily be found in literature [50,51], and therefore not all details are in-cluded here. Instead, some key properties of these techniques which have facilitated the experimental results to be obtained will be presented in this chapter.

3.1 Soft X-rays --- synchrotron radiation The first observation of synchrotron radiation was made at General Electric Research Laboratory in Schenectady, New York, in 1947 [52]. At first, the synchrotron radiation was seen as an undesirable energy loss when the latest technology of that time was applied for accelerating particles to velocities near to the speed of light. Later, researchers from all over the world started to build special circular storage rings only to produce high quality synchro-tron radiation. A synchrotron produces light with special qualities such as extreme brightness and short wavelength. Moreover, the photon energy can at certain conditions be varied continuously from IR to hard X-ray regime, with well-defined polarizations. These advantages can be fruitfully em-ployed in materials science at extremely high resolution.

A synchrotron radiation facility is mainly composed of an electron injec-tor, a storage ring, beam lines, and experimental stations. The linear accel-erator (injector) and the storage ring are employed to generate a beam of electrons, accelerate them, and keep them running with the same high speed in a circular orbit. At first, the synchrotron radiation was only generated by the electrons passing through the bending magnets installed between the straight sections of the storage ring. In order to get higher photon flux, inser-

38

tion devices called wigglers and undulators were later designed and installed in the straight sections of many storage rings around the world. All experi-ments presented in this thesis have been performed using radiation from an undulator.

A conventional undulator consists of a set of alternating dipole magnets, forcing the electrons to move in an oscillating trajectory with acceleration. The spectrum of radiation is discrete –consisting of the so-called harmonics –due to the constructive and destructive interference of the waves emitted at each point of the oscillating electron beam trajectory. The photon energy generated by the undulator can be defined from the following formula:

)2

1(2

222

2θγ

γλλ ++= K

nu

n

where n is the harmonic number, uλ is the periodic length of the magnets,

K is the undulator parameter which is determined by the magnetic field (the strengths of the magnets and the gap between them) and uλ . By setting the gap to different values, the photon energy can be varied. The radiation pro-duced by an undulator is delivered to the experimental station with the help of the so-called beamline. A beamline is as a rule a set of vacuum chambers with different optical elements such as mirrors, and diffraction gratings. The latter element is probably the most important one, since it allows us to choose the exact photon energy to be used and redirect away all the others from the broad spectral range of an undulator.

3.2 Experiments at beamline I 411, MAX-lab All the experiments except for the PAD study of water clusters included in this thesis were done at MAX-lab, located in Lund, Sweden. There are three storage rings inside the big experiment hall: MAX I, MAX II and MAX III (see figure 3.1). At the time of writing, the construction of the new advanced facility MAX IV is well under way in the north-east of Lund. In this work, we have only used MAX II, which belongs to the 3rd generation of the syn-chrotron radiation facilities. The electrons, at first generated in thermionic cathode are accelerated to 100 MeV. Then the electrons are boosted to 1.5 GeV by a LINAC- linear recirculation accelerator- before being injected into MAX II. The electron energy in the storage ring is 1.5 GeV with an emit-tance of less than 10 mrad, making MAX II well adapted for the soft x-ray radiation production. The insertion devices are installed in several straight sections of the ring. Several beamlines operate for the studies in various fields, such as surface science, atomic and molecular physics, condensed

39

matter physics, biology and chemistry, etc. The beamline I411 has been used in our study, and it will be described in the following paragraphs.

Figure 3.1 The MAX-lab synchrotron radiation facility. Figure is taken from ref.[53].

Beamline I411 is a soft X-ray beamline, which started operation in 1999. It was designed for many research fields, such as surface science, studies of gas-phase atoms and molecules, clusters, and experiments on liquids. The beamline is installed at the output of an undulator source. The undulator has 87 dipole magnets arranged in 43 periods. The period length of the undulator is 59 mm. The gap can be varied from 20 mm to 300 mm for covering the photon energy from 40 eV to 1500 eV.

Figure 3.2 The I411 soft x-ray undulator beamline. Figure is taken from ref.[54].

The figure above presents a schematic layout of beamline I411, which will now be described from the left to the right. At the output of the undulator there are horizontal and vertical baffles to confine the light passing though. It also helps to cut away the light at certain undesirable photon energies. The synchrotron-radiation beam then falls at a grazing angle at the front spherical mirror (M1), which focuses the beam slightly. The monochromator, which is a modified Zeiss SX700 device, contains two optical elements: a plane mir-ror (M2) and a diffraction grating (G1). The grating G1disperses the radia-tion and the plane elliptical mirror (M3) focuses the dispersed radiation onto the exit slit, and only the chosen monochromatic radiation passes through the

40

narrow slit. Further down the beamline there is a toroidal refocusing mirror (M4). There is about two-meter distance between M4 and the experimental end-station. After M4 there is a differential pumping stage which allows us to keep a pressure difference of several orders of magnitude between the end-station and the mirror chamber. Further down there is the so-called one-meter section, allowing external experimental stations to be installed.

The permanent experimental station consists of the ionization chamber with the electron energy analyzer, the preparation chamber, and the introduc-tion chamber. The mirror M4 creates a beam waist, making the beam dimen-sions acceptable both for the permanent end-station and in the 1 meter sec-tion [55,56]. The ionization chamber including the VG Scienta electron energy analyzer can rotate under high vacuum around the SR beam, making it pos-sible to detect the electrons at different angles to the horizontal polarization plane of the existing x-ray radiation. The electron energy analyzer installed at I411 is a Scienta R-4000 hemispherical electrostatic analyzer (figure 3.3). The electrons emitted from the sample as the result of ionization are acceler-ated or retarded by an electrostatic lens -to make them pass though the hemi-spherical electrodes of the analyzer at a certain fixed energy, the so-called pass energy. In front of the electrodes there is an entrance slit which defines the energy interval cut out for further analysis. Electrons of different kinetic energies are focused on different position of the detector. The detector in-corporates two 40mm diameter multichannel plates (MCP) and a phospho-rus-covered screen monitored by a CCD camera; the 2D detector image is seen in real time on the computer screen. The number of electrons imaged by the detector are counted and then sent to the computer. Finally, the energy distribution of the photoelectron can be recorded as a spectrum.

41

Figure 3.3. Schematics of a SCIENTA electron energy analyzer. The synchrotron radiation ionizes the cluster beam in the spectrometer focus point (35 mm below the conical lens cover seen on the drawing). The photoelectrons passing first through the lens and then through the hemispherical electrostatic field are imaged by the detec-tor. The figure is taken from ref.[54] .

3.3 Experiments at beamline UE 112 / PGM 1, BESSY II The PAD study of water clusters has been done at BESSY II, German syn-chrotron radiation facility in Berlin. Similar to the MAX II storage ring at MAX-lab, BESSY-II is also one of the third generation sources. The elec-trons at first generated in a thermionic cathode are accelerated to 70 keV (electron current of 300 mA). Then the electrons are boosted to 50 MeV (6 mA) before entering a smaller acceleration ring of 96 m circumference. Here, they are accelerated to 1.7 GeV and stored in a ring of 240 m circum-ference. Undulators and other insertion devices are also used at BESSY-II. The radiation provided by BESSY II can range from terahertz up to hard X-rays.

In principle, the PAD study could have also been done at MAX-lab, how-ever, it would have been much less convenient. The only soft x-ray gas-phase beamline at MAX-lab has an undulator delivering radiation which polarization is in the horizontal plane and it cannot be changed. There the rotatable end station makes it possible for the electron analyzer to measure the signal of the emitted photoelectrons at various angles. The PAD study of inert clusters has been done there [7], but technically such an experiment is rather demanding. Moreover, for production of water clusters, an in-vacuum tank with liquid water is used, and it is not convenient to rotate this tank.

42

Luckily the UE112-PGM-1 beamline at BESSY-II is equipped with an ellip-tically polarizing undulator which can also produce linearly polarized light in an arbitrary plane. In other words one can rotate the radiation polarization plane and have the cluster setup and the spectrometer fixed.

The beamline UE112-PGM-1 has an undulator of the APPLE-II type [41,57–59] with 32 periods, a period length of mm1120 =λ and a minimum bandgap of 22.2 mm. The photon energy can be ranged from 20 to 700 eV. The energy resolution of this beamline is fine with less than 1 meV for 100 eV photon energy. The magnet rows of apple-type undulator (Figure 3.4) can move with respect to each other, making it possible to choose the polariza-tion direction and even to produce circularly polarized radiation. The coils can be used to adjust the magnetic field in the emission point.

Figure 3.4. Schematic drawing of an APPLE-II undulator. The arrow in each mag-net block represents the magnetization direction. The figure is from Ref.[41]

The end-station we have used at beamline UE112-PGM-1 is not a permanent one. Users have to bring their own end-stations. For our experiment, we have also used a similar to the described above Scienta electron energy analyzer, which was placed at the “magic angle” relative to the horizontal plane. The water clusters were generated by a water cluster source similar to the one described in the next chapter.

43

4 Cluster/NPs production

In the works included in this thesis, several different kinds of clusters/NPs have been produced with various dedicated clusters sources. The sources for the salt cluster production are presented in the sub-chapters from 4.1 to 4.3, and at the end of chapter 4 there is a description of the setup for production of the metal-based NPs.

As briefly mentioned above, we have for the experiments on dry and wet alkali-halide clusters used: the adiabatic expansion cluster source (developed in-house first for inert gas or water clusters), which has been attached to the ionization chamber of the beamline end-station with the mounted on it ana-lyzer. The cluster source and the ionization chamber are separated by a skimmer so that the pressure in the ionization chamber can be kept suffi-ciently low for the detector of the analyzer. As mentioned above, a pick-up method has been employed in the study, which demands relatively big clus-ters of argon or water to be produced in the primary cluster source. The salt vapour to be picked-up by the primary cluster beam is created in an oven placed in the ionization chamber.

The metal/metal-oxide mixed composition NPs were created basically by the so-called gas-aggregation method using magnetron sputtering to produce the primary metal-atom vapour. The oxidation has been performed in two different ways, either by reactive sputtering or by “doping”, as described below in details.

44

4.1 Adiabatic expansion --- inert-gas cluster source

Figure 4.1. Schematics of the adiabatic expansion source. Figure is from ref.[3].

Figure 4.1 shows the schematics of the inert-gas cluster source. This setup has originally been used to study the size-dependent properties of inert-gas clusters. After the expansion, the clusters grow mainly due to the adsorption of free atoms. The clusters grow within the gas jet, and at some point, when the gas density has decreased sufficiently in the expansion, the growth proc-ess stops. In this work, as shown in the figure, high pressure Ar gas passes through the nozzle and enters the vacuum chamber. The nozzle is not just a microscopic hole but a long cone of a small opening angle diverging towards vacuum. In this work, a continuous Ar cluster beam has been used to pick up alkali-halide vapour and to form clusters out of it in a heat exchange process. Liquid nitrogen cooling is used to keep the nozzle at a cryogenic tempera-ture. In principle, a larger gas backing pressure and a lower temperature of the nozzle lead to larger clusters.

Hagena introduced the so-called expansion parameter Γ* [3], which can be calculated from the gas expansion conditions. By using the empirical relation between Γ* and <N>4, the mean size of Ar clusters can be estimated [60]. Our apparatus is capable of creating Ar clusters containing thousands of atoms.

45

4.2 Adiabatic expansion --- water-cluster source Several years ago the first photoelectron spectroscopy experiments on water clusters have been performed at the beamline I411 using a dedicated adia-batic-expansion source [28]. Different sizes of water clusters could be pro-duced. The principle of water cluster production is similar to that for the inert-gas clusters with the difference that liquid water is first heated in an oven to produce vapour. Free water clusters are created by supersonic ex-pansion of the water vapour, in some cases together with seeding the vapour helium through a conical nozzle. The nozzle is additionally heated to avoid local condensation.

In the works included in the thesis, the water clusters have been produced by two setups (at two different labs –MAX-lab and BESSY) with the same principle but not the same parameters. The water cluster source at MAX-lab used a nozzle of d = 150 µm diameter and α = 10° half opening angle. And at BESSY-II we used a nozzle of d = 80 µm diameter and α = 15° half open-ing angle. The free-water-cluster source used to produce the wet-salt clusters at MAX-lab has been a setup similar to the one in [13]. The description of wet cluster production will be given in the next section. Due to their similar-ity, only the cluster source used at BESSY-II will be described here -with the details as follows in this section. The size of water clusters can be deter-mined from the expansion parameters by a scaling law derived by Bobbert et. al [27,39]. The stagnation pressure p was derived as the vapour pressure of water at the reservoir temperature given. Scaling laws state a relation be-tween the parameters of a supersonic expansion used for cluster production and the mean cluster size produced in it. Although thermodynamically in-spired essentially they are empirical. The one used in our work has been suggested by Bobbert et al.[27] It reads:

a

n

q

cc

nqeq kTrT

TdpDN

=

−

1000

13

,

where a , q and D are empirical parameters which for water clusters were

optimized to values of 1.886, 0.634 and 11.6. cT = 5684 K and cr = 3.19 Å

are the characteristic temperature and characteristic radius for water, see

Ref.[27]. αtan/933.0: ddeq = is the ‘equivalent diameter’, k the Boltz-

mann constant. Within the given formalism, cluster sizes have a systematic

uncertainty of approx. 7 % from an inaccuracy of the thermocouple refer-

ence temperature, which influences the temperature measurement. There is also another experimental work on the clusters from which a lin-

ear relationship between the binding energy shift of a certain molecular or-

46

bital response in clusters and the cluster radius has been derived. The for-mula describing it is as follows:

3

−−Δ

Δ=bmono EEG

GN

Here, Emono and Eb denote the binding energy of the monomer peak and the observed cluster signal. GΔ is the Gibbs energy of solution. For the best agreement with the results, the value for GΔ of 1.4 eV is usually chosen [39].

4.3 The modified pick-up source As mentioned above, both the dry and the wet salt clusters in the study under discussion have been produced by the pick-up method. The pick-up source was originally developed to produce metal clusters [4]. The main compo-nents of the adiabatic expansion source have been the same as described above for the inert-gas and water cluster production. As mentioned above, in the latter case of wet clusters the cooled nozzle has been replaced by a wa-ter-oven with an integrated nozzle. The schematics of the present pick-up setup is shown in figure 4.2. The oven with the salt has been placed down-stream in the ionization chamber. This oven contains a separate removable crucible for the sample, which can be heated either resistively or inductively. For the salt studies, the crucible has been made out of stainless steel and has been resistively heated. A copper coil with circulating water is placed around the oven to reduce the radiative heat-load on the surroundings inside the chamber. There is an entrance and an exit hole in the oven to allow the Ar/water clusters pass through and pick-up the vapour. As discussed above, the skimmer between the adiabatic-expansion source and the ionization chamber plays an important role in keeping the pressure in the ionization chamber low enough for the experiment. It also cuts off the uncondensed atoms/molecules in the outer part of the cluster beam.

The alkali-halide clusters are formed as the water or Ar clusters pass through the oven towards its exit hole. The size of the salt clusters can be varied by changing the formation conditions: the size of Ar/water cluster and/or the vapour pressure of the heated species in the oven. In the works included in this thesis, we changed the size of the clusters formed in the pick-up process by varying the temperature of the oven, thereby changing the vapour pressure of the alkali-halide, and thus the amount of molecules encountered by the argon/water cluster during their passage through the oven.

47

Figure 4.2. Schematics of the pick-up setup. The figure is from ref.[4].

4.4 Gas aggregation --- magnetron sputtering-based source As mentioned above, all the metal-based NPs produced in the works in-cluded in this thesis were created by the gas-aggregation method, which implements magnetron sputtering for the primary atomic vapour production. In our cluster apparatus, the vapour can be also created by resistive heating of an oven containing the metal, however this vaporization method has not been implemented here, and is not discussed further in this work. The aggre-gation process takes place inside a liquid-nitrogen-cooled cylindrical cry-ostat. Magnetron sputtering of a solid material as a primary process makes the present work different from the classical cluster gas-aggregation ap-proaches developed, for example, in the groups of E. Recknagel [61] and T.P.Martin [6,22]. The first magnetron-based cluster source has been devel-oped by H. Haberland’s group [62]. In our source, at the exit of the cryostat there is a copper nozzle with about 20 mm long and 2 mm diameter channel. The so-called sputtering target represents a metal disc of 50 mm diameter and 6 mm thick. The target can be made either of one component or multiple components like the one schematically shown in Figure 4.3. Ar gas has been employed as the sputtering gas and also the carrier gas. The Ar gas was let in near the surface of the target. The metallic vapour is produced via the bom-bardment of the target surface by the Ar ions. The gas aggregation proceeds

48

via the vapour atoms’ and gas collisions with each other and subsequent condensation into NPs. For some experiments, He gas was also used to stabi-lize the sputtering and the flow through the cryostat. For both gases the input pressures are a few mbar. We establish empirically the values for both gases for each material being sputtered. The condensation length –the distance between the target and the exit of the nozzle has been ≈ 20 cm. The metal target can be sputtered using a DC power generator, and the insulator target has to be sputtered by RF plasma bombardment. The discharge power used for different materials is different. Here, we have sputtered only metals using DC plasma discharge. However, for some reactive metals, like Al, we needed RF power to remove the surface layer of oxide before the DC sput-tering could be started. The magnetron sputtering devices have been com-mercial products.

Figure 4.3 Schematics of the gas-aggregation source.

The magnetron sputtering-based gas-aggregation source can be used to pro-duce pure metal NPs and also the multi-component NPs. Figure 4.3, illus-trates the example of fabricating YbAl nanoalloy particles the study of which is presented in Paper VII. The target was clamped out of two parts: one of Al, the other of Yb. The mixed-composition NPs are formed by self-assembly, governed by thermo dynamics. In some cases the structure of mixed compo-sition NPs is dominated by a uniform mixture of constituent metals. How-ever, for some metals with large differences of properties, the radial segrega-tion can appear, or even the core-shell structure can be formed as we discuss in this work.

49

4.5 Reactive sputtering and doping methods of nanoparticle oxidation The NPs consisting of metal oxide are also a topic of broad current interest from both fundamental and applied perspectives. To produce this kind of NPs, one can, in principle, directly vaporize the oxide bulk material and subsequently force the vapour to condense into NPs. However, it is not easy to get high density of metal oxide molecules in the vapour phase compared to the pure metal atom vapour. Therefore, we have tried to fabricate the NPs by oxidizing the pure metal vapour produced first. The NPs can be oxidized at different steps of the aggregation process (see figure 4.4). We have used two different stages in the gas-aggregation process, where oxidation can lead to significantly different structure of NPs.

The first method of oxidation can be attempted on the already formed me-tallic NPs –by oxygen doping at the exit of the cryostat. In this case the pure metallic NPs are first produced according to the same procedure as described above. In our setup a hollow “doping” ring with multiple radial holes on the inner surface has been mounted perpendicular to the NP beam and just at the output of the nozzle. The oxygen pressure inside the doping ring could be regulated in a fine way by a high precision dozing valve and controlled by a gas-independent gauge. In this way, only rather reactive metals like Al and Yb but not Pb, Ag, Cu or Sn could be oxidized. The oxidation of NPs in the doping case has been shown to set on at the surface and then, with the oxy-gen dose increase, can penetrate into the inner part.

The second way to produce oxide-containing NPs is by the so-called reac-tive sputtering. In such a case the O2 gas is premixed with the sputtering gas Ar. This is a very efficient way to oxidize the metals that can not be oxidized by doping. The reason for this is probably the presence of excited, ionized and dissociated oxygen in the very vicinity of the target. Subsequently the oxide molecules and the pure metal atoms condense into the NPs. Earlier in our group, self-assembly of clusters out of several pairs of mixed inert gases has been studied using the adiabatic expansion source with the coexpansion of the gas mixture. Self-assembled metallic clusters have also been investi-gated earlier using our gas aggregation source in which the vapour from two metals –Na and K- was created by a resistively heated oven. In all the ex-periments, the high cohesive energy constituent elements have been found in the bulk of the mixed NPs. As for the reactive sputtering method, the ex-periments, in which the self assembly has been addressed for Pb, Sn, Yb, Cu, Ag, etc, and their oxides have been done, and the results on the Yb and Pb will be given in this thesis. The reactive sputtering has also been used to produce the nitrides [63].

50

Figure 4.4. Schematics of two different approaches to the oxidation process: in the upper figure it takes place before and in the lower figure after the aggregation.

51

5 Results

The Results Chapter of this thesis is divided into three parts: the first part will describe the study of alkali halide ionic-bonded clusters, both dry (Paper I, II) and wet ones (Paper III); the second part is the PAD-related study of water clusters (Paper IV) and Pb based mixed-composition NPs (Paper V); at last there will be a part describing the studies of the radial component distri-bution in metal/metal-oxide and alloy/alloy-oxide mixed composition NPs (Papers V, VI, and VII).

5.1 From dry to wet --- studies of alkali halide clusters As just mentioned above, first a study is presented on alkali-halide clusters existing in two phases: as “dry” clusters and as “wet” clusters. Dry clusters denote alkali-halide molecules bonded together in a nanoscale particle; wet clusters in this study mean alkali-halide molecules incorporated into water clusters. First, the background of this sub-field in the cluster research will be briefly presented. Then some additional (relative to those given above) de-tails of the methods of such cluster production and investigation will be de-scribed.

The work presented in the thesis can be seen in the context of a broader perspective: the experimental investigation of ionic compounds at nanoscale -exemplified at the first stage by dry salt clusters, and progressing to wet salt clusters. The study as a whole represents an investigation on the objects with increasing complexity in the composition and environment, but at the same time serving as close to ideal model systems. As the result a deeper under-standing of the electronic and geometric structure of alkali-halide clusters at nanoscale has been obtained. In other words, additional light has been shed on - the corresponding details of the interaction between the constituent at-oms and the surrounding neighbour atoms or molecules, which in the pre-sented works have been either alkali or halogen atoms or water molecules.

52

5.1.1 Ionic bonds in the initial and final states In complicated natural systems like sea-spray, alkali-halide clusters should exist with various sizes. This motivates the study of the size-dependent properties of these clusters not only from the fundamental science perspec-tive, but also from the environmental chemistry point of view. Using the pick-up method, we have been able to produce alkali-halide clusters of dif-ferent sizes.

In Paper I and Paper II, the photoelectron spectra of core- and valence-levels of the metal and halide ions of the alkali-halide clusters have been presented, and the chemical shifts in the binding energy relative to the corre-sponding monomer levels for both core and valence levels have been deter-mined. An electrostatic model (described above) has been constructed to calculate the core-level binding energy shifts in the clusters for various clus-ter sizes and model parameters. In this thesis, we will only show the results for NaCl and KCl clusters.

Na 2p core level spectra have been recorded at different clustering condi-tions (paper I). The NaCl dimer signal could be detected at ≈ 1 eV lower binding energy compared to the monomer signal. The cluster Na 2p core level feature has been observed to shift to ≈ 3 eV lower binding energy rela-tive to that of the monomer. For two experimental conditions (changed salt vapour concentration), different cluster peak positions have been observed (figure 5.1). We assign it as due to two different dominating cluster sizes created in each case. The lowest binding energy cluster features (figure 5.1, spectrum A) should correspond to the largest sizes as explained below.

For monomers, we have observed the Na 2p signal at 39.4 eV, which is accurately reproduced by the model calculation (see table 5.1). In order to use the electrostatic-model for clusters, we need several parameters, which are not exactly known: the interatomic distance, atom polarizabilities, cluster sizes, and the site of the atom giving the contribution to the experimental spectrum. To make reasonable predictions one can start with the parameters for the solid crystal. It should be mentioned here that the solid-state po-larizability, derived by fitting the results of an optical refraction experiment in 1953, has remained unquestioned since then, making its accuracy less trustworthy. However, in the absence of any better value, it has been imple-mented in the cluster energy calculations as the first approximation.

53

Figure 5.1. Photoelectron spectra in the Na 2p core-level region. Cluster spectrum A is recorded with a higher oven temperature creating higher salt-vapour density.

The Na 2p core-level binding energy of an atom at a specific position in clusters, such as corner, edge, or face has been calculated using the electrostatic model. It has shown a very weak size-dependence for sizes above 2*2*2 atoms. (As briefly mentioned above, it has been established in the earlier mass-spectroscopic studies [6,22] that alkali-halide clusters are formed as cuboids. In our model studies first cubic clusters have been treated). Let us consider, for example, the results for a cluster of the size 4*4*4 for which the energies are 37.8 eV (corner), 35.5 eV (edge) and 34.2 eV (face) for the three sites respectively. The higher-coordination sites appear at lower binding energies. Since the fraction of these sites increases with increasing cluster size, the overal cluster signal should shift towards lower binding energies (figure 5.1, spectrum A). Experimentally, the cluster responses are found in the region between 35.5 eV and 37.5 eV. Thus, despite of the relative crudeness of the model, we have got quite a good matching between the calculation and experiment.

Molecule Core-level Binding energy(eV) Calculated

energy(eV) Shifts ∆E(eV) (=Ecluster-Emonomer)

NaCl Na 2p 39.3 39.3 ≈2.8 to3.1

KCl K 3p 25.1 25.1 ≈1.8 to2.8

Table 5.1: The experimental and model binding energies for the metal ion in NaCl and KCl monomers and the experimental energy shifts from monomers to clusters.

54

The polarizability of Na is rather low compared to that of Cl. One way to cross-check the model and the role of the polarizability is to investigate clusters with one component fixed and the other changed. Such a test has been performed with different metal atoms and the same halogen. The atom with higher polarizabiltiy is potassium, and KCl clusters have been studied by us in the same way as those of NaCl, probing the K 3p core-level region.

As shown in Figure 5.2, the K 3p spectra in clusters also shift to lower binding energy with increasing oven temperature from A to B. As in the case of NaCl clusters, we assign this to increasing cluster size.

One should mention here that the K 3p shifts recorded for KCl clusters occurred to be smaller compared to those of Na 2p in NaCl clusters at similar conditions. This can be explained as due to the interatomic distance being considerably larger in solid KCl than in solid NaCl. The binding energies of the corner, edge and face atoms of KCl clusters have been also calculated using the electrostatic model for a range of cluster sizes and the interatomic distance and the polarizability of the solid [64]. As in the case of NaCl clusters, only a very weak size dependence of the site-specific energies has been disclosed for the clusters larger than 2*2*2 atoms. The corresponding model values for clusters with 4*4*4 atoms are 24.8 eV(corner), 22.2 eV(edge), and 21.3 eV(face). Again, the model reproduces the experimental results rather well. In figure 5.2 the peak maximum of the cluster response in the spectrum is close to the calculated binding energy for the edge atoms.

Molecule Core-level Binding

energy(eV) Calculated energy(eV)

Shifts ∆E(eV) (=Ecluster-Emonomer)

NaCl Cl 2p3/2 202.6 202.7 ≈1

Cl 2p1/2 204.2 204.2 KCl Cl 2p3/2 202.2 202.1 ≈0.7 Cl 2p1/2 203.7 203.7

Table 5.2: The experimental and model binding energies of Cl 2p level for NaCl and KCl monomers and the experimental energy shifts from monomers to clusters.

55

Figure 5.2. Photoelectron spectra in the K 3p core-level binding energy region. The temperature in the oven was increased gradually. The solid vertical blue lines at the bottom are the calculated results for different sites: from the left to the right- the corner, the edge, and the face.

The binding energy shifts for the halogens due to the transition from a monomer to a cluster have also been measured for NaCl and KCl. The corre-sponding photoelectron spectra are shown in figure 5.3. Remarkably, as mentioned yet above, the cluster responses of the Cl 2p core-level of NaCl and KCl clusters shift to higher binding energies relative to the monomer. This has been attributed to the disappearance of both the Coulomb interac-tion and a part of the polarization interaction in the final states.

The experimental and the calculated binding energies of Cl 2p core-level photoelectrons are presented in table 5.2. Both the experimental and calcu-lated shifts for NaCl clusters are larger than those of KCl clusters. Similarly to the metal atom case this is due to the larger interatomic distance of KCl solid relative to the NaCl solid.

56

Figure 5.3. Photoelectron spectra in the Cl 2p core-level binding energy region. The upper figure shows the response of the beam containing KCl clusters, dimers and monomers. The lower figure shows the Cl 2p level of the species in the beam with NaCl clusters.

In our works included in the thesis we have demonstrated that a simple model can be used to quantitatively explain the binding energy shifts for both NaCl and KCl clusters. The calculated energies are close to those ex-perimentally observed. As mentioned above, we have used the parameters for the solid alkali-halide salts [64]. If instead the somewhat larger (≈ 20%) polarizability value reported for Cl in free alkali-halide molecules [65] is used, the matching becomes slightly worse even for the molecule itself. However, for both published values of Cl polarizabilities in alkali-chlorides, the matching of the calculation to the experiment for monomers is much better than for clusters. The explanation of the discrepancies with the ex-periment can be the parameters used in the model: the interatomic distance and polarizability. We will now briefly describe our investigation of how

57

reasonable variations of these parameters affect the calculated binding ener-gies.

5.1.2 Halogen polarizability at nanoscale In order to investigate the role of the interatomic distance and polarizability for the clusters, model calculations have been performed with the parameters varying in the range ±40% of the solid value. The aim has been to optimize these parameters so the model calculation would match the experimental results better. While the metal halide interatomic distance is known to change when molecules form solids, and the values for both phases have been established with high precision, there are much fewer studies of the phase dependent behaviour of the polarizability. However there are at least some theoretical reports in which the decrease of polarizability (relative to the monomer value) was predicted already for the dimers of salt molecules [64,66–68]. Our model tests have been performed with the assumption that the atomic polarizability may decrease further in clusters (relative to dimers) in view of the larger electronic cloud compression for the constituents.

The calculated K 3p core-level electron binding energy for different clus-ter sites, distances, and polarizabilities for a cluster with 4*4*4 atoms are shown in figure 5.4 below:

Figure 5.4. The model values of K 3p core-level binding energies for the halogen polarizability around that of the monomer value.

The values for the solid interatomic distance and monomer polarizability from [64] are shown as vertical red lines. The interatomic distance for clus-

58

ters has been suggested to be between that of the molecule (the lower limit) and solid (the upper limit) [45]. In other words, the distance in clusters should not be larger than in the corresponding solid. Furthermore, the larger the distance is, the higher the metal core-level binding energy in alkali-halide salts is. In our model calculation, even if we use the largest possible value, i.e. the value of the solid, the calculated binding energies are still lower than those in the experiment. Thus the variation of the distance in the reasonable range does not remove the discrepancy, but some other effect must contrib-ute.

The effect of the metal polarizability on the K 3p core-level binding en-ergy has also been discussed in details in Paper II. It has been shown that a little change occurred when it was varied in a large range around the solid value. We ascribe this to the low polarizability of the metal ions relative to the halide ions, and the longer distance from a probed metal ion to the other metal ions. The last parameter to vary in the model is the halogen polariza-bility. If we use smaller polarizability, not only do the calculated binding energies shift to higher binding energy, but also the differences between the four sites become smaller (see figure 5.4). This would be consistent with our experimental results, where we get a wide peak with all the site responses overlapping. Therefore, it is reasonable to assume that the polarizability in alkali-halide clusters is even smaller than derived (≈ 60 years ago) for solid alkali-halide crystals from optical refraction experiments. The polarizability of Cl in the clusters is also lower than for the molecule [65] and that esti-mated for a dimer in [67]. This can be explained by the overlap of the elec-tron clouds what reduces their mobility.

Figure 5.5 shows a series of photoelectron spectra taken in the K 3p bind-ing energy region at different clustering conditions. This series has been obtained with the spectrometer acceptance axis at 90°to the polarization plane of the radiation. When the temperature in the oven is low (case A: the lowest spectrum), the cluster peak is located at a higher binding energy. If an “effective” polarizability for Cl of ≈2 Å3 is used, the case A position of the cluster feature is close to the calculated value for edge sites. This can be attributed to smaller clusters being formed at lower temperatures, since at smaller sizes it is natural to expect the relative intensity of the edge compo-nent to be larger than the intensity for the sites of higher coordination. As the oven temperature gradually increases, the intensity in the binding energy region higher than for the edge sites gradually grows, making the total clus-ter peak shifting to lower binding energy where the other cluster sites should be observed. The model values calculated for different cluster sites are pre-sented both in figure 5.4 (solid horizontal blue lines) and in figure 5.5 (as vertical lines at the bottom). The light-green rectangles from the left to right refer to edge, surface and bulk regions. The cluster feature position in top spectrum is close to the value calculated for the atoms at face-centers. This is consistent with an expectation that the relative intensity of face atoms will

59

gradually increase and finally will become the main contribution to the spec-trum. The bulk is weak and should appear in the even lower binding energy range. For the cluster size around 100 atoms, there is a very small bulk frac-tion and its signal is attenuated due to the finite electron escape depth.

Figure 5.5: Photoelectron spectra in the K 3p binding energy region recorded at 79.5eV photon energy and with the spectrometer at 90 degrees. The thin red vertical line indicates the position of the monomer signal. Position A corresponds to the calculated value for the cluster edge-atoms, and B for the face-center atoms using “effective” polarizability value and solid interatomic distance. The site propensity effect comes out in the spectra as the size of the clusters increased.

To summarize the results: the chloride polarizability deduced from the experimental results for the clusters is noticeably lower than that estimated for a separate chloride ion, or for an ion in a free alkali-halide molecule and dimer. Using the “effective” polarizability, the size- and site-dependent properties and their spectra manifestation can be well described by the elec-trostatic model.

60

5.1.3 Wet salt clusters --- solvation at nanoscale

Figure 5.6 Photoelectron spectra in the regions of metal core-levels in wet clusters for different alkali-halides. The responses of different wet salt clusters are observed at the same binding energy as in the studies of salt solutions [19].

In the literature, the studies of alkali-halides in aqueous solution by means of photoelectron spectroscopy on a liquid jet propagating in vacuum have been presented by at least two groups [19,69,70]. The core-level binding energies of alkali-metal and halogen ions solvated in water have been recorded at these conditions by B. Winter et. al [19] and G. Öhrwall et. al [70]. In the studies of wet alkali-halide crystals an increasing surface enrichment of highly polariz-able halide anions has been reported [71]. In the sense of salt concentration and aggregation phase wet clusters are different from the solutions studied, and depending on concentration could be closer to wet crystals. In the pre-sent study, wet salt clusters have been formed by alkali-halide molecules being picked up by relatively small water clusters. As discussed in the intro-duction, such particles resemble the sea-spray droplets existing in nature. To mimic the composition and chemistry in such droplets, using wet alkali-halide clusters is an important step towards the ultimate experiment dis-cussed in the introduction. In order to get information on the structure of wet salt clusters the binding energies of alkali metals and halogen atoms have been compared to the corresponding values for alkali-halide aqueous solu-tion and dry clusters.

61

The valence spectrum of water clusters with the size of a few hundreds of molecules has been also recorded in this set of experiments. In the present work it is ≈ 2 eV lower in binding energy compared to the monomer. For the alkali halides NaBr, NaCl, NaF, KBr, KCl, and KI samples the valence-region, metal core-region and halogen core-region spectra have been re-corded (figure 5.6). One of the observations to make from figure 5.6 is that, the Na 2p and K 3p core-level binding energies for different salts are rather close in wet clusters. For all wet clusters, the Na 2p binding energies have been found at around 35.5 eV, and for K 3p energies at around 22.4 eV. These values are practically the same as the energies measured for aqueous salt solutions but different from those measured for dry salt clusters. We interpret this as due to the dissociation of salt molecules into metal cations and halogen anions in water surroundings in clusters. The ions get solvated in a shell of water molecules, which makes their response similar to that in aqueous solutions studied in a liquid jet experiment.

A peculiarity to note here is that the total line width of the wet-salt cluster peaks for metal levels is 1.0±0.2 eV, which is considerably smaller than in the salt solution experiments (1.4±0.2 eV) [19]. This can be explained by a large charging effect, which broadens the spectral features in the experi-ments with the liquid jet. This also can make the binding energy values ex-tracted from such an experiment less accurate.

The binding energy shifts for halogen atoms have been also obtained for all wet salt clusters studied. In figure 5.7, the results for wet NaCl and KCl clusters are presented. The spectra have been calibrated using the molecular lines. Similar to the dry clusters, the response of the halogen atoms in clus-ters shifts towards higher binding energy. This can be explained by the anion interaction with the surroundings: the negatively charged halide ions become bound to water molecules in the wet salt clusters via charge-dipole interac-tion. The water molecule dipole has in this case both inherent and induced contributions. The charge and thus the larger part of the polarization energy disappear after the photoionization in the final states of halogen atoms. Qualitatively the effect is similar to what happens in the dry clusters, but quantitatively this case is different. The Cl 2p response for both salts in wet clusters has been observed at binding energies close to those of the aqueous solutions. The shift of Cl 2p response in the case of salt solvation in clusters has been larger than that for dry clusters.

The conclusion we may draw on the basis of the core-level binding en-ergy shifts for both the cations and anions of wet alkali-halide clusters is that at the present experimental conditions, the alkali halide molecules picked-up by water clusters dissolve in them and manifest themselves as ions sur-rounded by a solvation shell of water molecules in the same way as in aque-ous solutions.

62

Figure 5.7 Photoelectron spectra in the region of Cl 2p core-level for NaCl and KCl wet clusters. The spectrum contains also molecular response from the effusive beam of the oven with the salt.

5.2 Reduced anisotropy in the photoelectron distribution from clusters and NPs In this part I shall summarize our studies of photoelectron angular distribu-tion (PAD) for two substances: 1) the water clusters (Paper IV); and 2) Pb/Pb-Oxide mixed-composition NPs (Paper V).

5.2.1 Decrease of anisotropic parameters in water clusters As is well-established, liquid water consists of water molecules connected via hydrogen bonds, which constantly break and reform [13,72,73]. However, the details of the interaction between molecules in the liquid are still far from being completely clear [74]. As an attempt to improve the understanding of water as a substance, the water clusters have been produced and studied in the frames of this work. As has been just discussed above, at MAX-lab, we

63

mainly studied the behaviour of alkali halide solvation in water by picking up alkali halide molecule with the water clusters. Then after that, a further and a more specific study of water clusters has been made in collaboration with Dr. U. Hergenhahn’s group using the BESSY-II synchrotron facility in Berlin, Germany. As described in chapter 3, the PADs of different orbitals in water clusters have been studied with a special type of undulator [41,57,58], which can provide linearly polarized radiation with a rotatable polarization plane. As, again, briefly mentioned earlier in the thesis, this makes it con-venient to record the angular dependence of the photoelectron signal: one can have the cluster source and the electron spectrometer fixed and just ro-tates the plane of linear polarization of the synchrotron radiation. This method is relatively novel. As discussed above, the PAD shows a directional propensity with respect to the polarization plane. The exact value of the ani-sotropy parameter β depends on the intrinsic response of the ionized orbital to the applied field, and (for molecules and clusters) on scattering effects which the continuum electron experiences in the potential of the ionized particle. This is why, in principle, PADs contain information both on geome-try and electronic structure of the free NPs under investigation.

Already in the 1980s the PAD of isolated water molecules has been stud-ied, and the β anisotropy parameters reported, by Truesdale et al. [75] and Banna et al. [76]. Winter et al. studied the PAD of liquid water and compared their results to the gas-phase β parameters in their analysis. To the best of our knowledge, no PAD study on water clusters has been done by any other group. First, in the present work, the PAD of a water monomer has been measured with photon energy of 40 eV and 60 eV (Paper IV). The gas phase monomers were produced by lowering both the stagnation pressure of the reservoir and the nozzle temperature until no spectral features with lower binding energies than the gas phase molecular adiabatic 1b1 peak were seen. The results have been compared to the previous studies [75,76], and this com-parison showed that the anisotropy parameters derived for three valence orbitals were very close to the earlier published values. The β parameters measured by us agree with the literature data within the respective error bars.

In Paper IV, we have also presented the results of PAD studies on water clusters. The main observations have been similar to the previous studies of the inert-gas clusters obtained by several groups [7,77,78], however, there have also been some differences. The authors [77,78] have concluded that the de-crease of β for the inert gases in the transition from free atoms to the atomic clusters have been caused by elastic scattering of the outgoing photoelec-trons (i.e. a final state effect), and that the orbital-specific effects, such as band formation did not play a role for the decrease of β. However, as our measurements have shown, in the water clusters, not only the elastic scatter-ing effect, but also some not unambiguously identified intrinsic effects were responsible for the decrease of the anisotropy in PAD.

64

In the works included in this thesis, three sets of the PAD values of water clusters have been presented, one at 40 eV and two at 60 eV for the orbitals 1b1, 3a1 and 1b2 (see figure 5.8). The difficulty in this study is that the sig-nals from the monomers and clusters overlap, and only for the 1b1 orbital, the signals can be well separated. This also allows us to determine the rela-tive molecule-to-cluster intensity. For the other orbitals we have assumed that the cluster intensity fraction is the same as for the 1b1 orbital. In the experiments, we can only determine the β parameter analyzing the area un-der the signal containing both the cluster and the monomer responses for each orbital. We have partitioned the measured β parameter according to a certain rule derived above in Chapter 2: β = cβcl + (1 – c)βm, where sub-scripts cl and m distinguish the cluster and monomer components. The size of the clusters was determined by using the scaling law [27]. We have ob-tained mean cluster sizes of N = 58 for the 40 eV data set, and N = 84 for both 60 eV data sets.

Figure 5.8 (a): Photoelectron spectra of a mixture of water clusters and water monomers, recorded at a photon energy of 60 eV at an angle θ = 0° and at the ‘magic angle’ of 54.7° (dotted and solid trace correspondingly). (b): The β parame-ter for different orbitals derived from the experiments. The Roman numbers corre-spond to the different partitions made.

In our study a dramatic decrease of anisotropy parameters for all the valence orbitals in water clusters have been clearly observed. We can start with the HOMO level, the resolved cluster contribution for 1b1 channel, which can be spectroscopically separated from the monomer photoelectron spectrum. Photoemission from the cluster 1b1 level is much more isotropic than from its molecular counterpart. The right panel of figure 5.8 shows the corre-sponding PAD values we have got for the water clusters with the size around 84 and recorded at 60 eV. The values are shown for different mole-cule/cluster signal partitions. The left panel shows the partly overlapping spectra from the clusters and monomers measured at the magic angle and at 0 degrees. The differences between the two spectra can be clearly observed.

65

As mentioned above, compared to the study of inert gases, the results for water are not surprising. As described in the literature and mentioned above, the decrease of anisotropy parameters for clusters has been explained by the intracluster electron elastic scattering. The possibility of intercluster scatter-ing in a low-density beam is very weak and can therefore be neglected. The previously published works on the subject [49,72,78] have only presented quali-tative studies of the scattering effect in the final state. Therefore, a simple model to quantitatively calculate the scattering effect can be useful to con-struct, and can be especially important for the more complex substance – water clusters. The details of the model have been given in the theoretical part in Chapter 2 of this thesis. The isolated monomer differential cross-section has been taken from Cho et al. [79]. The structure of water clusters used in the model is a cuboid of 4*4*5 water molecules, with an O-O dis-tance of 3 Å. The results of the model-based calculations compared with the experimental results are given in the table below. Orbital Photon energy βcl(exp) βcl(model) β(mol.) 1b1 40 eV 0.83(8) 0.91(4) 1.38(8) 3a1 0.73(16) 0.70(3) 1.08(8) 1b2 0.42(16) 0.49(2) 0.75(12) 1b1 60 eV 1.17(8) 1.31(3) 1.59(8) 3a1 0.99(12) 1.15(3) 1.41(8) 1b2 0.70(18) 0.85(2) 1.04(12)

Table 5.3. Angular distribution parameter β for the outer valence orbitals of water clusters, both the experimental and calculated model values, and in the last row for the molecule.

The results of the model calculation clearly reveal the decrease of the anisot-ropy parameters. However this decrease caused by scattering is not always enough to explain the large experimental difference between separate mole-cules and clusters. In five out of six cases in Table 5.3, the β values we measured are lower than the ones derived from the scattering model. While for the 40 eV data, the difference is still within the experimental error bars, for two of the three 60 eV data points the measured data differ by more than one standard deviation.

Therefore, in this study, we cautiously interpret the results as the evidence that there are some intrinsic differences between molecular and cluster PADs, and not only caused by the final state electron scattering. Among the most probable reasons could be the changes taking place in the two least strongly bound molecular orbitals, which are significantly modified by hy-drogen bonding in clusters.

Apart from other conclusions our results also clearly show that, in the photoelectron spectroscopy related to water, the photoelectron angular dis-tribution effects have to be taken into account.

66

5.2.2 Elastic scattering effects for metal bulk atoms In Paper V, the PAD has also been studied for the NPs which consist of a metal and its oxide. As mentioned above, PADs contain information both on geometry and electronic structure of the probed sample. The aim of the study has been to shed some light on the geometric structure of multi-component NPs by PAD, namely on the distribution of different substances in such NPs. This approach has been applied to the Pb/Pb-oxide NPs. As discussed in Chapter 3, PAD studies can also be carried out at the I411 beamline at MAX-lab, although the polarization plane of this beamline is fixed. The rotatable ionization chamber of the end station makes it possible to measure the photoelectron signals at different angles. In this subsection, we first demonstrate our results on PADs for two energy levels for pure Pb NPs, and the determination of component distribution in Pb/Pb-oxide NPs will be given in the following sub-section.

As mentioned above, the Pb NPs were produced by the gas-aggregation method based on the magnetron sputtering. The details of the NP production have been described in Chapter 4. The ionization of the Pb 5d core level and the valence level has been studied at two different angles relative to the hori-zontal polarization plane, namely at the “magic angle” and 90 degrees. Typi-cal spectra are shown in figure 5.9.

Figure.5.9 Photoelectron spectra of Pb 5d5/2 (left panel) and valence (right panel) levels for metallic Pb clusters, the dashed-line spectra have been recorded at the “magic angle” and the solid-line ones at 90°. The valence spectra have been normal-ized to the lower-energy 6p3/2-related part of the valence band.

The 5d response contains both bulk and surface contributions separated by ≈ 0.2 eV. The change of the observation angle can, in principle emphasize either bulk or surface responses, what would manifest itself in the change of

67

the spectral shape of the corresponding feature. As seen from figure 5.9, the 5d lineshape for both angles is very similar. That is also because the anisot-ropy parameter for Pb 5d is quite small (≈ 0.5) at 50 eV photon energy. Thus the distribution of electrons is rather isotropic from the very moment of ioni-zation, and is thus similar for photoelectrons escaping from the bulk and from the surface. As for the valence level, it is not easy to calibrate the inten-sity of the spectra unambiguously, however, a clear change of the spectral shape can be observed. In principle, we should not observe the valence re-sponse at 90°, since the angular anisotropy parameter β at 40 eV photon energy is close to 2.0 [80], meaning that there are hardly any electrons ejected at 90° with respect to the polarization plane. There should thus be practically no signal at all if we measured the valence level for the separate Pb atoms. Therefore, we can conclude here that in the case of clusters, the effective β for the same level is considerably lower than for separate atoms, mainly due to the elastic scattering of the photoelectrons coming out from the cluster interior up to the surface.

It is well known that elastic scattering effects are much weaker for the photoemission from surface atoms than from bulk atoms [7]. The PAD of the surface should be more atomic like, with the valence electrons mainly com-ing out from the clusters parallel to the polarization electric vector, and hence cannot be detected by the spectrometer at 90°. We will use this situa-tion in the next sub-section.

From both PAD studies of water clusters and Pb NPs, we can come to a conclusion that, the PAD anisotropy parameters in clusters are clearly re-duced in comparison to their counterpart in separate atoms and molecules. The course of this reduction is to a large extent elastic scattering but there could be some other, intrinsic effects, as band formation for the valence.

5.3 Core-shell structure of metal/metal oxide NPs In this section, the results on the geometric structure of NPs consisting of metal and metal oxide will be presented. This structure has been deduced from different peculiarities of the NPs spectral behaviour. The first sub-section describes how, using PAD the core-shell structure of Pb/Pb-oxide NPs has been disclosed (Paper V). The other two sub-sections, describe how the structure of Yb-based NPs was deduced by analyzing the relative inten-sity changes and binding energy shifts of the related spectral features. The analysis shows that NPs consisting of Yb and its oxide can be produced with the reversed radial order of components: there can be Yb/YbO or YbO/Yb core-shell structured NPs (Paper VI). Moreover, in tri-component NPs con-taining Yb, Al, and metal-oxide the YbO/Yb/Al “sandwich-like” NPs (Paper VII) have been produced and studied.

68

5.3.1 Disclosing core-shell structure of multi-component NPs using photoelectron angular distribution As mentioned above, NPs consisting of group IV elements or/and their ox-ides are intensively studied as possible building-blocks for novel devices [42,81–86]. Our studies show one of the possibilities to create such NPs with core-shell structure thus providing radial segregation of conduct-ing/semiconducting properties. As mentioned in Chapter 4, the direct oxida-tion of Pb NPs by exposing their metallic surface to oxygen flow (doping method) occurred to be not efficient. To reach the aim, mixed Pb/Pb-oxide NPs have been produced by a method based on so-called reactive sputtering. Oxygen mixed with the sputtering Ar gas has been let into the cryostat, and oxidation of Pb atoms occurred very close to the target, forming oxide mole-cules already before the aggregation.

The arrangement of elements in the NPs when they were formed by ag-gregation/self-assembling process is one of our interests in this work. As mentioned above, the bi-component NaK NPs created by a similar method have been studied by Tchaplyguine et al.11, and the results show that the high cohesive energy element has been placed in the interior of the core-shell NPs. In Papers V and VI, the mixed-composition NPs have only one metal, and the second substance is its oxide. The aim of the study has been to shed some light on what kind of arrangement was formed and how different ar-rangements could appear when the NPs were created by the self-assembly process.

The amount of oxygen admixed into argon and then let in the cryostat has been precisely controlled, and the corresponding changes of spectra for dif-ferent oxidation conditions have been unambiguously recorded. For Pb NPs, starting from a certain fraction of oxygen admixed in argon, the spectra con-tain both Pb and Pb-oxide signals. As the oxygen fraction increases the rela-tive intensity of Pb-oxide increases. At certain “ultimate conditions”, only the oxide feature appeared in the spectra. In the work included in this thesis, special attention is given to the structure of the mixed NPs at the condition when both the metallic Pb and Pb-oxide responses are seen in the 5d spectra of the NPs. In figure 5.10, the 5d5/2 and the valence spectra recorded for such a case at the “magic” angle and at 90° are presented. In the left part we can see that for both angles, the Pb and Pb-oxide responses are present, and the relative intensity of metallic signal in the spectrum recorded at 90° is higher than in the other recorded at the magic angle. However, when comparing the valence level responses at different angles (right panel), for the spectrum recorded at 90°, no metallic signal can be observed where it should appear. This means that electrons from the NP metallic part are highly anisotropic, and it is the electrons from the surface which preserve their anisotropy –in contrast to the bulk. Therefore, from the observation above, the structure of the mixed NPs should be as follows. In the mixed NPs, the metallic Pb at-

69

oms are present only on the surface. The clusters, born first as agglomerates of oxide molecules, are swept downstream, into the regions where oxidation is less likely, but where the Pb atomic vapour, getting colder and colder fur-ther away from the target, is still present. This kinetics scenario can indeed lead to the core-shell structures with the oxide core and metallic surface.

Figure.5.10 Photoelectron spectra of Pb 5d5/2 (left panel) and valence (right panel) level for Pb/PbO mixed-composition NPs, the dashed-line spectra have been re-corded at the “magic” angle, and the solid-line ones at 90°.

Apart from the specific kinetics of the cluster self-assembly in our source, it is also the thermodynamics which is behind such a segregated distribution of components. In such a gas-aggregation cluster source a certain mobility in-side the clusters is preserved within the formation time, so the lowest-energy cluster stoichiometry can be achieved. It is energetically more favorable to place a constituent with weaker interatomic bonds on the surface of the NPs, while the one with stronger bonds -in the bulk. The competition between the metal and metal oxide surface energies related to the bond strengths and also the interface energy minimization which plays a role in the cluster formation process favor the same segregated structure. This competition defines to a great extent whether there is a well-separated shell of one of the subcompo-nents or a gradual gradient of their concentration. And as our study has shown, there has been a clear separation between two components –lead and its oxide.

70

5.3.2 A@B and B@A core-shell NPs formed of Yb and YbO As mentioned above, to produce core-shell NPs out of a metal and its oxide, direct “doping” of the NP surface by oxygen was successful in some cases, but not for all metals. For the metals we used in the works included in the thesis, only the reactive metals like Yb and Al could be oxidized by doping. Yb occurred to be one of the metals that could be both oxidized by reactive sputtering and by doping using our setup.

In this part of the study, the preparation and characterization of Yb/YbO NPs created via two different methods are described. The first method is doping, in which the pure metal NPs have been produced first, and the oxy-gen was used at a later stage to oxidize primarily their surface. The second method implemented reactive sputtering, in the same way as for the Pb/Pb-oxide mixed composition NPs. As discussed above, this allows the formation of core-shell NPs with the oxide in the bulk. The photoelectron spectra at both conditions have been recorded and two characteristic spectra are pre-sented in figure 5.11 (the red vertical traces). The spectra give us the first impression of the differences of the electronic structures of the NPs for dif-ferent production methods. For both methods, the oxygen was stepwise let into the reaction volume, and photoelectron spectra from the beam contain-ing NPs have been recorded at each condition.

For the doping case, at lower oxygen fractions, the most apparent obser-vation is the 0.2 eV up shift of the bulk Yb 4f response. We have attributed the shift to the formation of oxide on the surface, and the new peak position as coming from Yb atoms at the interface between the metallic bulk and the oxidized surface. As the amount of oxygen increases to a higher level, the bulk feature decreases until no signal at its energy position can be detected. The surface oxide signal increases and shifts to lower binding energy. Even in the case of the strongest exposure to oxygen no apparent intensity has been detected in the binding energy region where the tri-valent Yb oxide would manifest itself. We conclude that the doping results in a core-shell structure with a divalent YbO shell surrounding a metallic Yb core.

The NPs with reversed radial structure (relative to that of the doping case) were produced by reactive sputtering, in which the oxidation process took place prior to the NP aggregation. As mentioned above, the reactive sputter-ing is an efficient method to oxidize metals due to the presence of dissoci-ated, ionized and excited oxygen in the vicinity of the magnetron target. The cryostat, where the aggregation takes place, is 30 cm long with a condensa-tion path of ≈ 20 cm. The concentration of charged particles decreases dras-tically with the distance from the target, so the formation of oxide molecules becomes much less likely. Therefore, further away from the target the prob-ability of metal atoms condensing on the oxide core becomes significant. In the experiment, the amount of O2 in the O2/Ar mixture in present experi-ments with Yb has been step-wise increased to around 5% and, as mentioned

71

above, each situation has been characterized by photoelectron spectroscopy. The most notable changes in the spectra take place in the region to the lower binding energy side from 4f5/2 bulk-peak position of the pure-metal NP – between ≈ 5.3 and ≈ 5 eV. This can be explained by the presence of a new peak related to the bulk 4f5/2 level. If so, there should be a corresponding second spin-orbit component to the lower energy side of 4f7/2 bulk peak. And indeed, the latter peak has a shoulder in the expected region. This extra dou-blet we attributed to be the bulk oxide, which assignment is based on two observations: 1) the reactive sputtering favors the oxidation at an early stage of NP formation; 2) and in the doping experiments Yb surface monoxide has been shown to be at a higher binding energy. We conclude that with a high probability, NPs with oxide in the bulk and pure metal on the surface were created by reactive sputtering.

Figure.5.11 The illustration of the binding energy positions for the responses of different sites in Yb/YbO mixed composition NPs. The spectra of NPs created by two different oxidation methods are vertically placed on both sides of the binding energy axis.

In this section, we have introduced our two methods of producing NPs with two different distributions of the same elements. For both cases we obtain the distribution close to the so-called core-shell geometric structures, but of the opposite radial orders.

The metal covered by its oxide is common, however, the reversed case is rather peculiar. Such a geometry becomes possible due to a combination of thermodynamic and kinetic conditions realized in the gas-aggregation method with reactive sputtering.

72

5.3.3 Core-shell bi-metallic nanoalloy YbAl As stated above, multi-component NPs are the subject of the main interest of this thesis. The materials based on this kind of NPs are already applied in nanotechnology and some are promising for different fields. NPs containing one-two elements with one or both oxidized are potential building blocks for complex functional nanomaterials. In this section and the section follow, we will describe the production of bi-metallic nanoalloy particles and also the oxidation of them. One of the aims of this study has been to make the NPs with the same components as used in a family of solid state lasers with the active medium based on garnets doped by rare-earth elements. As a combi-nation practically favorable to study, aluminum, ytterbium and oxygen have been chosen. Corresponding NPs containing these three elements have been produced and their electronic and geometric structure has been studied.

As mentioned above, the first step of the study has been to create the na-noalloy particles, which in this work have been produced by the gas aggrega-tion method based on magnetron sputtering. As again not once mentioned above, bi-metallic NPs were earlier produced by our gas-aggregation source [11], in which the mixture of two vapours was created by the resistive heating of two metals –sodium and potassium – in an oven placed inside the cryostat. In the work included in the thesis, the mixed vapour of two metals was also produced in the cryostat however by magnetron sputtering. To reach this aim, the two metals can be sputtered separately or can also be sputtered in a bi-metallic target as in this study. The target was a 50 mm diameter, 6 mm thickness disc consisting of two parts clamped together: one out of ytterbium and one out of aluminum. The concentration of the two metals can be varied using different fractions of each metal in the target. In this work, Yb was ≈ 40% of the target volume, and correspondingly aluminum ≈ 60%. However the fractions in the vapour phase were different due to different sputtering efficiencies.

To determine the structure of the mixed YbAl nanolloy particles, photo-electron spectra in both Yb 4f (see figure 5.12) and Al 2p (see figure 5.13) core-level regions have been recorded. Figure 5.12 shows the Yb 4f spectra for the pure Yb NPs and for the YbAl alloy particles. Instead of the four clear, well resolved peaks of pure Yb NPs, the spectrum for the alloy shows three peaks, what we explain as due to the shift downwards of the bulk spin-orbit components. One can also notice that the feature with highest binding energy of the alloy NPs is at the same binding energy and has a similar width as the surface 4f5/2 feature of pure Yb case. Therefore we assign this feature to the pure Yb surface. Similar spectral shapes- with three instead of four peaks- have been observed in the studies of ytterbium deposited on metallic substrates. R. Nyholm et al. [87] observed a similar spectral shape after a deposition of a thin Yb layer on Al monocrystal. N. Mårtensson et al. [88] have found that for two completed Yb monolayers on Mo surface the 4f

73

binding energy of Mo/Yb interface had shown up at about 0.5 eV lower binding energy than the pure Yb bulk. Therefore, we attributed the shifted, lower peak due to mainly the Yb atoms above the YbAl interface and to a smaller extent- to some bulk Yb atoms. The interface feature of YbAl na-noalloy is noticeably broader than the pure bulk feature as it should be due to the more complex Yb environment. There is another peculiarity observed in the Yb 4f spectra of nanoalloys: similarly to the case of ytterbium oxide discussed in the previous sub-section, there is only divalent Yb 4f response recorded in the YbAl alloy. The trivalent Yb would appear as a certain set of narrow spectral lines at higher binding energy than the divalent Yb. This is a strong support that Yb and Al have not been homogenously mixed in the NPs as Yb in a well-mixed alloy is tri-valent [89].

Figure 5.12 Photoelectron spectra of pure Yb nanoparticles (solid red line) and YbAl nanoalloy (dotted line) produced by our gas-aggregation source with magne-tron sputtering. The main difference is in the bulk response.

In order to verify our hypothesis of a pure-Yb surface on top of a core con-taining mostly Al, the Al 2p spectrum has also been recorded (see figure 5.13). The Al 2p binding energy of YbAl is observed to shift down by ≈ 2 eV compared to the pure Al NPs. This shift is close to the difference of work functions between Yb and Al macroscopic metals. Such an observation is consistent with the mixed composition NPs, in which Yb covers the surface (and reduces the work-function) and Al atoms are placed in the bulk and, in principle, possibly partly mixed with Yb deep inside the core. However, the

74

analysis of the Yb 4f spectra of the nanoalloy, does not support the mixing. We conclude that we can produce core-shell structured bi-metallic NPs with Al in the core and Yb or the surface.

The observed surface segregation of Yb is consistent with kinetic and thermodynamic considerations: when particles are formed inside the cryostat there is enough atomic mobility in them to reach the lowest-energy configu-ration in which Yb –the metal with lower cohesive energy- is pushed to the surface.

Figure 5.13 Photoelectron spectra of nanoparticles in the Al 2p binding energy region for pure Al NPs (obtained by us in a separate experiment not presented here) and YbAl nanoalloy particles.

5.3.4 Oxidation of nanoalloy YbAl particles --- the sandwich NPs Following the study in the last section, the oxidation of YbAl core-shell structured NPs has been attempted by the doping method, the details of which have been described in Chapter 4 in general and above in sub-section 5.3.2 for YbO preparation. Briefly, the YbAl particles have been oxidized by exposing the beam of the preformed YbAl NPs to a concentric flow of oxy-gen.

A series of measurements has been done with the amount of O2 used for doping stepwise increased. Similarly to the case with the oxidation of pure Yb NPs, a ≈ 0.2 eV upshift of the bulk/interface peak was observed at low oxidation degree with a doping pressure of ≈ 1 mbar (figure 5.14, spectrum

75

b). At low and intermediate oxidation (spectra b and c), the 4f5/2 surface re-sponse, which does not overlap with the other features in the spectrum and for the non-oxidized alloy has a well defined maximum at ≈ 6 eV, becomes broad, flat and without a clear maximum, what all bears a witness of more than one peak contributing to it. When the oxygen doping pressure increases, the central peak at ≈ 5 eV and the lowest energy peak (Fig. 5.14), which are due the Yb-Al interface 4f5/2 and 4f7/2 responses loose in intensity relative to the surface response. This intensity loss we can attribute to the coverage of the surface by the oxide, making the interface even deeper. In the ultimate oxidation case, the interface feature is still distinguishable in spectrum e but with relatively low intensity and shifted towards lower binding energy –relative to the non-oxidized case. Moreover, it is shifted below the interface response of the low/intermediate-oxidation case b. This is similar to the oxi-dized Yb in Yb/YbO NPs where the bulk oxide showed up at energies lower than those of the metallic bulk.

The spectra in the Al 2p region have been separately measured for the oxidized Al NPs (see figure 5.13). The Al 2p feature of the oxide was ob-served at about 2 eV higher binding energy than Al 2p feature of metallic Al. In the studies of YbAl nanoalloy particle oxidation, the intensity of Al 2p occurred to be rather weak with low signal to noise ratio. As we have shown in our earlier studies [90] on metallic Al NPs their bulk signal is responsible for the lower binding energy part of the Al 2p spectral feature. Therefore, it is tempting to assign some physical meaning to the lack of intensity on the lower energy side of Al 2p signal in the oxidized case. This observation is consistent with the conclusion above: the inner part of NPs, in this case the aluminium core, becomes buried even deeper when the oxide layer is formed on the surface.

At the intermediate oxidation conditions, the “sandwich-like” structure of YbO/Yb/Al seems to be formed. When the oxygen doping pressure is suffi-ciently high, the deepest Yb layer at the interface to the aluminium core is likely to become also oxidized, and a core-shell structure of Al-YbO be-comes the case.

76

Figure 5.14 Recorded in the Yb 4f region and up to 8 eV above it photoelectron spectra of YbAl alloy NPs oxidized by the “doping” method. From the bottom to the top, the oxygen doping pressure increases. The gaseous O2 signal with well-resolved vibrational lines is marked by a vertical light blue stripe, and the Yb 4f7/2 interface region in non-oxidized NPs (case a) is shown with a vertical light green stripe. The region where the intensity appears in the ultimate oxidation case e is marked with light yellow band. This intensity we assign to the Yb alloy bulk oxide.

77

6. Summary & Outlook

Studies of various multi-component clusters/NPs with different bonding types have been presented in this thesis. For the ionic-bonding clusters, both dry and wet alkali-halide clusters have been produced by a pick-up source, and studied by photoelectron spectroscopy using synchrotron radiation. The ultimate goal of this work could be seen in the investigation of compounds and processes relevant for the atmospheric chemistry. Gradually the com-plexity of the conditions in the experiments performed within the present work has been increased. In the first step dry alkali-halide clusters have been studied. The core-level binding energies of metal and halogen atoms in dry clusters of different sizes were measured: Significant negative shifts relative to the monomer signal have been observed for metal atoms, and positive shifts for halogen atoms. The experimental shifts for metal atoms have been shown to be about 3 times larger than those for the halogen atoms in NaCl and KCl clusters. Based on previous studies, the structure of alkali-halide clusters has been assumed to be cubic or cuboidal. On this assumption an electrostatic model has been constructed (paper I) accounting for the changes in the electron binding energies in the ionization process. Both the “charge-charge” Coulomb interaction and polarization interaction with all the atoms in the clusters have been taken into account in the model. The model calcula-tion has given results matching the experimental binding energies well. Us-ing the model results, the significant differences in binding energy observed with changing cluster size have been interpreted as due to the differences in the site specific responses. In the process of the work on the data on various salt clusters a deeper understanding of such systems has been gained. As an initial approximation, the polarizabilities used in the electrostatic model were the values published more than 50 years ago for salt molecules and for solid salts. Even though the calculated results using these values have been reasonable, there have been indications in several publications that the po-larizabilities in clusters could be significantly smaller than in molecules. As clusters are an intermediate phase between molecules and solids, the parame-ters could be different from both the molecular and solid cases. The effects caused by the changes in the interatomic distance, and in the polarizabilities of both cations and anions have been studied by model calculations and compared to the experiments on KCl clusters (Paper II). The interatomic distance in the alkali-halide clusters has been varied between the molecular and the solid limiting values, and its variation was shown to be unable to

78

improve the matching between the experiment and the model. The metal atom polarizabilities have been shown to only marginally affect the binding energies. Finally an “effective” electronic polarizability for chlorine in clus-ters has been introduced and estimated by optimizing the model calculation to fit the experimental results. The polarizability has shown to be smaller than that reported for separate molecules, dimers and even for the solid (pa-per II). By using the new “effective” halogen polarizability, it became possi-ble to describe more accurately and assign a clear physical meaning to the experimental response of dry clusters: The shape of the total cluster peak in the spectra changes from being an “edge-atom-rich” signal to “face-atom-rich” as the size of the clusters gradually grows. The bulk signal has hardly been observed even for the largest clusters what was not surprising in the view of the cluster size –below or around 100 atoms.

In the second step, the particles formed as the result of interaction be-tween alkali-halide vapour and water clusters have been studied (paper III). The core-level electron binding energies of both the alkali metal atoms and halogen atoms have been measured. Similarly to the case of dry clusters the negative binding energy shifts for metal atoms and the positive shifts for halogen atoms have been observed. The shifts for both metal and halogen atoms in wet clusters have been shown to be nearly the same as those for aqueous salt solutions. This can be seen as a proof of the alkali-halide mole-cules being dissolved in water clusters.

The PAD for water clusters has been studied in frames of this thesis. A clear reduction of anisotropy parameters for all three valence orbitals has been observed and explained mainly by photoelectron elastic scattering. A simple model has been built to calculate the effect of elastic scattering. The results show, however, that not only the intracluster scattering has lead to the decrease of the anisotropy of the distribution, but also some still unknown intrinsic effect could play a role in that. The anisotropy in the PAD for the Pb-oxide NPs has also been experimentally observed. This data has been used to shed light on the structure of Pb/Pb-oxide mixed-composition NPs.

Following the study of Pb-based NPs, the Yb/YbO mixed-composition NPs have been created with variable radial arrangement of the components and these particles have been investigated by photoelectron spectroscopy. Two production methods utilizing different oxidation order, before or after aggregation, have been used to create the reversed structure of the mixed composition NPs. From the photoelectron spectra measured in the Yb 4f binding energy region, the different structures of both cases could be de-duced.

In the last part of the thesis, we present our study of the most complex multi-component NPs studied by us, the YbAl nanoalloy particles and the oxidation of them. For this alloy NPs, the spectra in both Yb 4f and Al 2p binding energy regions have been measured. Analysis of the spectral features in comparison to the pure Yb and pure Al NPs, allows us to obtain a strong

79

evidence for that we have created core-shell structured NPs with Al in the interior and Yb on the surface of the nanoalloy particles. Step-wise oxidation has been performed and the corresponding spectra from the beam of NPs have been in-situ measured. At low oxygen exposure the oxidation occurs only to the surface monolayer of Yb, and in this case, we get “sandwich-like” NPs. As the amount of oxygen increases, the oxidation penetrates deeper into the NPs, and can reach the interface of YbAl. Oxidation of the bulk Al is not observed -likely due to its deep location in the core.

As a continuation of the project with wet alkali-halides the interaction be-tween even smaller water clusters and salt molecules is of interest to study. This would allow studying solvation at otherwise unattainably high concen-trations, mimicking clusters formed by water evaporation from sea spray droplets. Phenomena to be explored include phase separation at concentra-tions above the solubility limit, and surface propensity effects induced by ion-ion interaction which may contribute to the increased importance of bromide in atmospheric chemistry [69].

Another direction possible to follow is the studies of some other dry and wet salt clusters relevant for the atmospheric chemistry, such as sulfates. Analogously to the present work, the electronic and geometric structure of the clusters can be studied. However, here some additional technical devel-opment would be necessary, since most relevant salts, like ammonia sulfate, are easily decomposed by heating.

As for the metal-based NPs, it would be interesting to study their structure by other techniques like microscopy. For that a deposition and in-vacuum transportation setups should be developed.

On could also try to fabricate more complex and practically useful struc-tures under the gas-phase conditions. For example, doping of metal NPs by some organic molecule can be attempted, or reactive sputtering with other gases admixed into argon, like nitrogen or hydrogen sulfide is of potential interest.

81

Sammanfattning på svenska

Kluster / Nanopartiklar Kluster/ nanopartiklar syftar här på mycket små materieklumpar bestående av ”ett fåtal”, två till miljontals, atomer. Även om deras storlek är mellan molekylen och det fasta makroskopiska materialet, kan egenskaperna vara helt annorlunda än de hos motsvarande molekyl eller fasta material. Låt oss till exempel betrakta en bit av guld. Om vi kan mala den till ett mycket fint pulver, kommer bland annat färgen att ändras. Om vi maler guldet till storle-kar på några hundra nanometer, kan färgen bli blå, och för storlekar mindre än 100 nm blir färgen röd. Många andra märkliga egenskaper hos nanopar-tiklar har också tillämpats på olika områden. Genom att till exempel ersätta materialet i solceller och batterier med nanopartiklar har effektiviteten ökats väsentligt. Vissa andra nanopatiklar kan injiceras i blodet, vilket hjälper lä-karen att bota fler sjukdomar. I de studier jag presenterar i avhandlingen, har olika typer av kluster / nanopartiklar producerats i laboratoriet.

I atmosfären nära havsytan, finns en hel del våta partiklar av nanostorlek. Ett rimligt antagande är att dessa partiklar förutom vatten även innehåller andra ämnen från havsvattnet. Eftersom en del av partiklarna stiger till högre skikt av atmosfären, kommer vattnet förångas och lämna partiklarna, som blir torrare. Storleken på partiklar i atmosfären kan vara från några få till hundratals nanometer. Laboratorieframställda partiklar som är modeller av sådana atmosfäriska partiklar utgör den första partikelgruppen som vi stude-rar i denna avhandling.

En annan typ av partiklar som vi skulle vilja studera är metall-baserade nanopartiklar. Partiklarna kan bestå av endast ett ämne eller av flera kompo-nenter, och kan vara ren metall eller metalloxider. På grund av luftförore-ningar kan denna typ av nanopartiklar också hittas i atmosfären i städer och industriella områden. I de flesta fall är de skadliga för människors hälsa. Dock har allt två sidor, och denna typ av partiklar kan också vara bra, till exempel i nya material.

För att bäst förstå de partikeltyper vi nämnde ovan, är det bästa sättet att samla in dem eller tillverka dem i laboratoriet, och sedan studera dem med hjälp av vetenskapliga metoder.

82

Elektronisk struktur och fotoelektronspektroskopi Ett materials egenskaper beror på de kemiska element som det innehåller, distribution av dem, och även materialets geometriska struktur. Olika mikro-skopimetoder kan direkt observera materialets geometriska struktur. I denna avhandling studerar vi den elektroniska strukturen, som ger andra sätt att bestämma ett materials struktur och egenskaper. För en enskild atom, kan det finnas många elektroner i olika orbitaler runt kärnan. Elektronerna nära kär-nan är de elektroner är starkast bundna, medan de yttersta elektronerna är lättare att avlägsna. Därför behöver vi mer energi för att slå ut elektroner nära kärnan. Den elektroniska strukturen hos den enskilda atomen kan be-skrivas som elektroner i diskreta tillstånd av olika energi. När tittar på ett system med ett fåtal atomer, kan de diskreta tillstånden kvarstå, dock kan de tillståndet av samma grundämne variera beroende på krafter mellan atomer-na och andra effekter. Allt som antalet atomer ökar, börjar elektrontillstån-den att överlappa och så småningom närma sig en band-liknande struktur som för ett fast material. Från gapet mellan valensbandet och ledningsbandet kan vi bedöma konduktiviteten hos den fasta substansen. För de klus-ter/nanopartiklar som vi studerat i denna avhandling, innehåller den elektro-niska strukturen information om skillnaderna mellan nanopartiklarna och den enskilda atomen respektive det fasta materialet. Den elektroniska struk-turen hos atomer på olika platser i nanopartiklarna kan vara mycket olika. I nanopartiklar bestående av flera komponenter, kan tillstånden förskjutas mot högre eller lägre energi på grund av olika typer av bindning mellan atomer-na. Till exempel, när en metall oxideras, kan den elektroniska strukturen hos metallen förändras dramatiskt. Därför är det mycket viktigt att hitta ett sätt att kartlägga den elektroniska strukturen hos atomer, molekyler, kluster och fasta ämnen. Och i denna uppsats använder vi metoden som kallas fotoelek-tronspektroskopi.

För att mäta den elektroniska strukturen, behöver vi en ljuskälla, provet, och en elektronenergianalysator. Ljuskällan vi använt är synkrotronljus. Dess främsta fördelar är det mycket höga fotonflödet med avstämbar våg-längd, från UV-ljus, över mjuk till hård röntgenstrålning. Strålningen kan, om fotonenergin är tillräckligt hög, slå ut elektroner ur atomerna. Detta är den så kallade fotoelektriska effekten. Dock betyder det inte fotonenergin måste vara så hög som möjligt. Elektronerna i olika orbitaler har maximal sannolikhet att slås ut vid olika fotonenergier. Med hjälp av elektronenergi-analysatorn kan den kinetiska energin hos de emitterade elektronerna mätas, och deras intensitet som funktion av kinetisk energi presenteras som en kur-va som kallas fotoelektronspektrum. Genom att titta på spektrum, kan vi observera en förändring av den elektroniska strukturen i materialet vid olika förhållanden.

83

Resultat vi fick För alkalihalid kluster, har alkalihaliden värmts i en ugn, där en mycket tät ånga kan produceras. Nästa steg är att sända en mycket kall stråle av ädel-gaskluster genom ångan, där ett ädelgaskluster kan plocka upp ett flertal alkalihalidmolekyler, och därigenom bilda alkalihalidkluster. För de våta klustren är produktionsprocessen mycket liknande; i stället för den kalla ädelgasklusterstrålen användes en vattenklusterstråle.

Fotoelektronspektra för alkalihalidkluster mättes under olika experimen-tella förhållanden. Genom den modell vi byggt, verifierar vi idén att alkali-halidklustren är kuboid med en storlek av cirka 1 nm. Genom att jämföra alkalihalidklustren med motsvarande molekyl och fasta material, fann vi att klustrens strukturparametrar är olika, och varierar med storleken. Till exem-pel kan halidjonernas polarisabiliteter vara olika för joner på olika platser i klustret. För de våta klustren, fann vi att alkalihaliderna är solvatiserade re-dan i nanovattenkluster eftersom deras fotoelektronspektra är desamma som för vattenlösningar av alkalihalider.

Fotoelektroner avges inte lika mycket i alla riktningar från klustren, utan uppvisar en vinkelfördelning. Om en atom i ett kluster bestrålas med synkro-tronstrålning, kan de emitterade fotoelektronerna spridas på grannatomer innan de kommer ut ur klustret. I detta fall kan fotoelektronernas vinkelför-delning ändras, och detta fenomen kan också användas för analys av klust-ren. Vi har mätt fotoelektronvinkelfördelning av vattenkluster och bly/blyoxid multi-komponentnanopartiklar. För vattenkluster fann vi att vin-kelfördelning är mer isotrop än för enskilda molekyler. För blybaserade na-nopartiklar har denna metod använts för att bestämma den radiella fördel-ningen av de ingående ämnena. Resultatet visar att ren bly har finns på ytan av bly/blyoxid multi-komponentnanopartiklar.

Nanopartiklar bestående av en metall och dess oxider är intressanta äm-nen. I fallet med blybaserade nanopartiklar, har vi funnit metall på ytan, vilket inte är vanligt i makroskopisk skala, där metallbulken alltid är täckt av sin oxid. I denna studie försöker vi också att producera en nanopartikel med omvänd struktur, det vill säga med metalloxiden på ytan av metallen. Detta har förverkligats i fallet med ytterbium/ ytterbiumoxid multi-komponentnanopartiklar nanopartiklar; vi kan framställa dem med två olika, inbördes omvända, strukturer med antingen metall på ytan eller i det inre.

Andra komplexa system har också studerats; vi har producerat nanoleger-ingspartiklar bestående av två olika metaller. I detta fall måste vi producera dessa två metallångor i samma volym, och låta dem själv-aggregera inuti. Från de fotoelektronspektra vi mätte, har tydliga radiella fördelningar av de två olika metallerna observerats. För till exempel nanolegeringar av alumini-um och ytterbium fann vi Yb atomerna på ytan. Den radiella fördelningen av de två olika komponenterna beror främst på deras storlek och kohesivenergi. Ännu mer komplexa system också har prövats, genom att vi också oxidera

84

ytan av nanolegeringspartiklarna. Så slutligen, kan vi producera "sandwich-liknande" nanopartiklar med ytterbiumoxid på ytan, aluminium innerst och ytterbium emellan.

85

Bibliography

1. Hongbo Shi, Ruth Magaye, Vincent Castranova, and Jinshun Zhao, Particle and Fibre Toxicology 10, 15 (2013).

2. Shuwen Zeng, Xia Yu, Wing-Cheung Law, Yating Zhang, Rui Hu, Xuan-Quyen Dinh, Ho-Pui Ho, and Ken-Tye Yong, Sensors and Actuators B: Chemical 176, 1128–1133 (2013).

3. Olle Björneholm, Gunnar Öhrwall, and Maxim Tchaplyguine, Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment 601, 161–181 (2009).

4. S. Osmekhin, M. Tchaplyguine, M. -H. Mikkelä, M. Huttula, T. Andersson, O. Björneholm, and S. Aksela, Physical Review A 81, 1–5 (2010).

5. M. Tchaplyguine, R. R. Marinho, M. Gisselbrecht, J. Schulz, N. Mårtensson, S. L. Sorensen, A. Naves de Brito, R. Feifel, G. Ohrwall, M. Lundwall, S. Svensson, and O. Björneholm, The Journal of Chemical Physics 120, 345–56 (2004).

6. T. P. Martin, The Journal of Chemical Physics 69, 2036 (1978). 7. G. Öhrwall, M. Tchaplyguine, M. Gisselbrecht, M. Lundwall, R. Feifel, T.

Rander, J. Schulz, R. R.T. Marinho, A. Lindgren, S. L. Sorensen, S. Svensson, and O. Björneholm, Journal of Physics B: Atomic, Molecular and Optical Physics 36, 3937–3949 (2003).

8. Tiberiu Arion, Melanie Mucke, Marko Förstel, Alex M Bradshaw, and Uwe Hergenhahn, The Journal of Chemical Physics 134, 074306 (2011).

9. Marko Förstel, Melanie Mucke, Tiberiu Arion, Toralf Lischke, Silko Barth, Volker Ulrich, Gunnar Öhrwall, Olle Björneholm, Uwe Hergenhahn, and Alex Bradshaw, Physical Review B 82, 1–6 (2010).

10. M. Tchaplyguine, S. Peredkov, A. Rosso, I. Bradeanu, G. Öhrwall, S Legendre, S. Sorensen, N. Mårtensson, S. Svensson, and O. Björneholm, Journal of Electron Spectroscopy and Related Phenomena 166-167, 38–44 (2008).

11. M. Tchaplyguine, S. Legendre, A. Rosso, I. Bradeanu, G. Öhrwall, S. Canton, T. Andersson, N. Mårtensson, S. Svensson, and O. Björneholm, Physical Review B 80, 9–12 (2009).

12. Juan Salafranca, Jaume Gazquez, Nicolás Pérez, Amílcar Labarta, Sokrates T Pantelides, Stephen J Pennycook, Xavier Batlle, and Maria Varela, Nano Letters 12, 2499–503 (2012).

13. O. Bjorneholm, F. Federmann, S. Kakar, and T. Moller, The Journal of Chemical Physics 111, 546 (1999).

14. Chunhui Li, Qingbo Meng, and Bo Brummerstedt Iversen, Physical Chemistry Chemical Physics 97, 95–97 (2013).

15. Christopher J Serpell, James Cookson, Dogan Ozkaya, and Paul D Beer, Nature Chemistry 3, 478–83 (2011).

16. M. Willander, O. Nur, Q. X. Zhao, L. L. Yang, M. Lorenz, B. Q. Cao, J Zúñiga Pérez, C. Czekalla, G. Zimmermann, M. Grundmann, A. Bakin, A. Behrends, M. Al-Suleiman, A. El-Shaer, A. Che Mofor, B. Postels, A. Waag, N. Boukos,

86

A. Travlos, H. S. Kwack, J. Guinard, and D. Le Si Dang, Nanotechnology 20, 332001 (2009).

17. Nam-Hyuk Cho, Taek-Chin Cheong, Ji Hyun Min, Jun Hua Wu, Sang Jin Lee, Daehong Kim, Jae-Seong Yang, Sanguk Kim, Young Keun Kim, and Seung-Yong Seong, Nature Nanotechnology 6, 675–82 (2011).

18. Wandared Pokapanich, Henrik Bergersen, Ioana L Bradeanu, Ricardo R T Marinho, Andreas Lindblad, Sebastien Legendre, Aldana Rosso, Svante Svensson, Olle Björneholm, Maxim Tchaplyguine, Gunnar Ohrwall, Nikolai V Kryzhevoi, and Lorenz S Cederbaum, Journal of the American Chemical Society 131, 7264–71 (2009).

19. Bernd Winter, Ramona Weber, Ingolf V Hertel, Manfred Faubel, Pavel Jungwirth, Eric C Brown, and Stephen E Bradforth, Journal of the American Chemical Society 127, 7203–14 (2005).

20. JE Campana, TM Barlak, RJ Colton, JJ DeCorpo, JR Wyatt, and BI Dunlap, Physical Review Letters 47, 1046–1049 (1981).

21. Svante Svensson, Journal of Physics B: Atomic, Molecular and Optical Physics 38, S821–S838 (2005).

22. T. Martin, Physics Reports 95, 167–199 (1983). 23. T. P. Martin, U. Zimmermann, N. Malinowski, U. Naher, S. Frank, F. Tast, and

K. Wirth, Surface Review and Letters 3, 281–286 (1996). 24. T. Moller, A. R. B. de Castro, K. Von Haeften, A. Kolmakov, T. Laarmann, O.

Löfken, C. Nowak, F. Picucci, M. Riedler, C. Rienecker, A. Wark, and M. Wolff, Journal of Electron Spectroscopy and Related Phenomena 101-103, 185–191 (1999).

25. C Bobbert and C.P. Schulz, The European Physical Journal D 16, 95–97 (2001). 26. Udo Buck, Ingo Ettischer, Mario Melzer, Victoria Buch, and Joanna Sadlej,

Physical Review Letters 80, 2578–2581 (1998). 27. C. Bobbert, S. Schütte, C. Steinbach, and U. Buck, The European Physical

Journal D 19, 183–192 (2002). 28. G. Ohrwall, R. F. Fink, M. Tchaplyguine, L. Ojamäe, M. Lundwall, R. R. T.

Marinho, A. Naves de Brito, S. L. Sorensen, M. Gisselbrecht, R. Feifel, T. Rander, A. Lindblad, J. Schulz, L. J. Saethre, N. Mårtensson, S. Svensson, and O. Björneholm, The Journal of Chemical Physics 123, 054310 (2005).

29. Bernd Winter, Emad F Aziz, Uwe Hergenhahn, Manfred Faubel, and Ingolf V Hertel, The Journal of Chemical Physics 126, 124504 (2007).

30. Sinhué López, a. Romero, J. Mejía-López, J. Mazo-Zuluaga, and J. Restrepo, Physical Review B 80, 1–10 (2009).

31. Yugang Sun and Younan Xia, Science (New York, N.Y.) 298, 2176–9 (2002). 32. M. -H. Mikkelä, M. Tchaplyguine, K. Jänkälä, T. Andersson, C. Zhang, O.

Björneholm, and M. Huttula, The European Physical Journal D 64, 347–352 (2011).

33. T. Andersson, C. Zhang, A. Rosso, I. Bradeanu, S. Legendre, S. E. Canton, M. Tchaplyguine, G. Öhrwall, S. L. Sorensen, S. Svensson, N. Mårtensson, and O. Björneholm, The Journal of Chemical Physics 134, 094511 (2011).

34. B. O. Dabbousi, J. Rodriguez-Viejo, F. V. Mikulec, J. R. Heine, H. Mattoussi, R. Ober, K. F. Jensen, and M. G. Bawendi, The Journal of Physical Chemistry B 101, 9463–9475 (1997).

35. C. A. Pope III, JAMA: The Journal of the American Medical Association 287, 1132–1141 (2002).

36. C. Nowak, C. Rienecker, A. Kolmakov, and O. Löfken, J, Journal of Electron Spectroscopy and Related Phenomena 101-103, 199–203 (1999).

87

37. S. Urpelainen, M. Tchaplyguine, M.-H. Mikkelä, K. Kooser, T. Andersson, C. Zhang, E. Kukk, O. Björneholm, and M. Huttula, Physical Review B 87, 035411 (2013).

38. Daniel B Miracle, Nature Materials 3, 697–702 (2004). 39. Silko Barth, Milan Oncák, Volker Ulrich, Melanie Mucke, Toralf Lischke, Petr

Slavícek, and Uwe Hergenhahn, The Journal of Physical Chemistry. A 113, 13519–27 (2009).

40. G. Öhrwall, M. Tchaplyguine, M. Lundwall, R. Feifel, H. Bergersen, T. Rander, A. Lindblad, J. Schulz, S. Peredkov, S. Barth, S. Marburger, U. Hergenhahn, S. Svensson, and O. Björneholm, Physical Review Letters 93, 1–4 (2004).

41. Shigemi Sasaki, Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment 347, 83–86 (1994).

42. V. Senz, T. Fischer, P. Oelßner, J. Tiggesbäumker, J. Stanzel, C. Bostedt, H. Thomas, M. Schöffler, L. Foucar, M. Martins, J. Neville, M. Neeb, Th. Möller, W. Wurth, E. Rühl, R. Dörner, H. Schmidt-Böcking, W. Eberhardt, G. Ganteför, R. Treusch, P. Radcliffe, and K.-H. Meiwes-Broer, Physical Review Letters 102, 3–6 (2009).

43. U. Hergenhahn, A. Kolmakov, M. Riedler, A. R.B de Castro, O. Löfken, and T. Möller, Chemical Physics Letters 351, 235–241 (2002).

44. D. Rolles, H. Zhang, Z. Pesic, J. Bozek, and N. Berrah, Chemical Physics Letters 468, 148–152 (2009).

45. G. Yalovega, A. V. Soldatov, M. Riedler, Mark R. Pederson, A. Kolmakov, C. Nowak, and T. Moller, Chemical Physics Letters 356, 23–28 (2002).

46. A. W. Potts and W. C. Price, Physica Scripta 16, 191–196 (1977). 47. Robert Hudgins, Philippe Dugourd, Jason Tenenbaum, and Martin Jarrold,

Physical Review Letters 78, 4213–4216 (1997). 48. H. Derenbach and V. Schmidt, Journal of Physics B: Atomic and Molecular

Physics 17, 83–93 (1984). 49. H. Zhang, D. Rolles, J. D. Bozek, and N. Berrah, Journal of Physics B: Atomic,

Molecular and Optical Physics 42, 105103 (2009). 50. F. J. Himpsel, Brazilian Journal of Physics 23, 31–52 (1993). 51. V. Schmidt, Electron Spectrometry of Atoms Using Synchrotron Radiation,

Cambridge Monographs on Atomic, Molecular and Chemical Physics (Cambridge University Press, Cambridge, 1997).

52. F. Elder, A. Gurewitsch, R. Langmuir, and H. Pollock, Physical Review 71, 829–830 (1947).

53. M. Lundwall, Rare-gas Clusters Studied by Electron Spectroscopy, Uppsala University, 2007.

54. Sonia Rondon, Surface Science Spectra 5, 97 (1998). 55. M. Bässler, J-.O. Forsell, O. Björneholm, R. Feifel, M. Jurvansuu, S. Aksela, S.

Sundin, S.L. Sorensen, R. Nyholm, and A. Ausmees, Journal of Electron Spectroscopy and Related Phenomena 101-103, 953–957 (1999).

56. M. Bässler, Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment 469, 382–393 (2001).

57. U. Englisch, H. Rossner, H. Maletta, J. Bahrdt, S. Sasaki, F. Senf, K.J.S Sawhney, and W. Gudat, Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment 467-468, 541–544 (2001).

58. J. Bahrdt, W. Frentrup, A. Gaupp, M. Scheer, W. Gudat, G. Ingold, and S. Sasaki, Nuclear Instruments and Methods in Physics Research Section A:

88

Accelerators, Spectrometers, Detectors and Associated Equipment 467-468, 21–29 (2001).

59. Kai Godehusen, Hans-Christoph Mertins, Tobias Richter, Peter Zimmermann, and Michael Martins, Physical Review A 68, 1–4 (2003).

60. R. Karnbach, M. Joppien, J. Stapelfeldt, J. Wormer, and T. Moller, Review of Scientific Instruments 64, 2838 (1993).

61. K. Sattler, J. Mühlbach, and E. Recknagel, Physical Review Letters 45, 821–824 (1980).

62. Hellmut Haberland, Journal of Vacuum Science & Technology A: Vacuum, Surfaces, and Films 12, 2925 (1994).

63. Marc-Alexandre Dubois and Paul Muralt, Applied Physics Letters 74, 3032 (1999).

64. Jack Tessman, A. Kahn, and William Shockley, Physical Review 92, 890–895 (1953).

65. A. W. Potts, T. A. Williams, and W. C. Price, Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences 341, 147–161 (1974).

66. H. W. de Wijn, The Journal of Chemical Physics 44, 810 (1966). 67. R. S. Chauhan, S. C. Sharma, S. B. Sharma, and B. S. Sharma, The Journal of

Chemical Physics 95, 4397 (1991). 68. PH. Citrin and T. Darrah Thomas, The Jouranl of Chemical Physics 57, 4446–

4461 (1972). 69. Niklas Ottosson, Jan Heyda, Erik Wernersson, Wandared Pokapanich, Svante

Svensson, Bernd Winter, Gunnar Ohrwall, Pavel Jungwirth, and Olle Björneholm, Physical Chemistry Chemical Physics : PCCP 12, 10693–700 (2010).

70. G. Ohrwall, N. Ottosson, W. Pokapanich, S. Legendre, S. Svensson, and O. Björneholm, The Journal of Physical Chemistry. B 114, 17057–61 (2010).

71. Sutapa Ghosal, Matthew a Brown, Hendrik Bluhm, Maria J Krisch, Miquel Salmeron, Pavel Jungwirth, and John C Hemminger, The Journal of Physical Chemistry. A 112, 12378–84 (2008).

72. B. Winter, R. Weber, W. Widdra, M. Dittmar, M. Faubel, and I. V. Hertel, The Journal of Physical Chemistry A 108, 2625–2632 (2004).

73. Bernd Winter, Emad F Aziz, Uwe Hergenhahn, Manfred Faubel, and Ingolf V Hertel, Optik 124504, (2011).

74. J.-H. Guo, Y. Luo, a. Augustsson, J.-E. Rubensson, C. Såthe, H. Ågren, H. Siegbahn, and J. Nordgren, Physical Review Letters 89, 1–4 (2002).

75. C. M Truesdale, S. Southworth, P. H. Kobrin, D. W. Lindle, G. Thornton, and D. A. Shirley, The Journal of Chemical Physics 76, 860–865 (1982).

76. M. S. Banna, B. H. McQuaide, R. Malutzki, and V. Schmidt, The Journal of Chemical Physics 84, 4739–4744 (1986).

77. H. Zhang, D. Rolles, Z. D. Pešić, J. D. Bozek, and N. Berrah, Physical Review A 78, 1–7 (2008).

78. D. Rolles, Z. D. Pesic, H. Zhang, R. C. Bilodeau, J. D. Bozek, and N. Berrah, Journal of Physics: Conference Series 88, 012003 (2007).

79. H. Cho, Y. S. Park, H. Tanaka, and S. J. Buckman, Journal of Physics B: Atomic, Molecular and Optical Physics 37, 625–634 (2004).

80. Manfred Krause, Patrick Gerard, Anders Fahlman, Thomas Carlson, and Agneta Svensson, Physical Review A 33, 3146–3150 (1986).

81. Hua-Jin Zhai, Jens Döbler, Joachim Sauer, and Lai-Sheng Wang, Journal of the American Chemical Society 129, 13270–6 (2007).

82. Transition Metal and Oxide Catalysts, 6136–6148 (2002).

89

83. Robert E. Benfield, Journal of the Chemical Society, Faraday Transactions 88, 1107 (1992).

84. K. A. Zemski, D. R. Justes, and A. W. Castleman, The Journal of Physical Chemistry B 106, 6136–6148 (2002).

85. Publisher Taylor and I Pollini, Philosophical Magazine 37–41 (2009). 86. Baolin Wang, Jijun Zhao, Xiaoshuang Chen, Daning Shi, and Guanghou Wang,

Physical Review A 71, 1–7 (2005). 87. Ralf Nyholm, Ib Chorkendorff, and Jenss Schmidtmay, Surface Science 143,

177–187 (1984). 88. N. Mårtensson, A. Stenborg, Q. Björneholm, A. Nilsson, and J. Andersen,

Physical Review Letters 60, 1731–1734 (1988). 89. En-Jin Cho, S.-J. Oh, C.G. Olson, J.-S. Kang, R.O. Anderson, L.Z. Liu, J.H.

Park, and J.W. Allen, Physica B: Condensed Matter 186-188, 70–73 (1993). 90. Tomas Andersson, Chaofan Zhang, Maxim Tchaplyguine, Svante Svensson,

Nils Mårtensson, and Olle Björneholm, The Journal of Chemical Physics 136, 204504 (2012).

90

91

Acknowledgments

This work has been performed with the help of many people. We have worked together in the experiment measurements, data analyzing and manu-script writing. First of all, I would like to express my deepest gratitude to my supervisors, Prof. Olle Bjorneholm, Doc. Maxim Tchaplyguine, and Prof. Svante Svensson. Olle gave me a lot of comments and ideas in the studies. And support me to go to a lot of places all over the world for conferences and experiments during 4 years, even some of them are not relevant to my current project. However, I got a lot of help from all the travels. Maxim is a very good supervisor, very patient and also very experienced, we are to-gether for the most of the time in my 4 year PhD studies. I have learnt a lot from him and he never said “no” to me and always let me try what I want. When I made mistakes, he has never condemned me but encouraged me. Without Svante, I think I would not be able to come to here, and all the other things would not happen. Svante is very busy, however, he always tried to talk to me and guild me when we met. I like listening his stories about his researches and wonderful life. I would also like to thank Dr. Uwe hergen-hahn, we have made a beautiful work together. Before met him, I even did not know what water molecule orbitals are. Prof. Xiaojun Xu is my supervi-sor for the master’s study, I like to discuss with him on the projects, and we have also continue working together after I came here. And also Prof. Zhejin Liu, the head of our group in China, I want to thank you for supporting me a lot in the study.

And I also would like to express my appreciation to my colleagues for help-ing me. Tomas helped me a lot not only in the work but also in a lot of things in the life here, especially helped me to deal with the documents that wrote in Swedish. Gunnar is very nice person to answer my very simple questions very patiently. I would also like to thank the colleagues Nils, Hans, Johan, Niklas, Wandared, Josephina, Yixiao and Melanie at the Department of Physics and Astronomy, Uppsala University; Stacey, Mathieu, Joachim, at the Department of Synchrotron Radiation Research, Lund University; Marko, Mikko, Leena, Kari, Lauri, Dmytro at Department of Physics, University of Oulu; Bernd, Marko, Stephan, and Isaak at BESSY II in Germany; Marc, Rajash at Soliel in France and Yuran, Qiushi, Samuli and Prof. Lindau at MAX-lab. Thank you all for doing experiments together and also the won-derful discussions we had.

92

Great thanks go to my previous colleagues at National University of Defense Technology in China, Prof. Tan is my former supervisor for the diploma work, he always encouraged me and helped me, I can never forgot he waited me at the entrance of hotel for two hours in a raining night when I came back China. And other colleagues in my previous group, Hongyan, Zhoupu, Jia-jian, Hanwei, Xiaolin, Rongtao, Haotong, Yanbin, Luguang, Chenhan, Hankai, Xiaoming, Gelun, Zhihong, Jianfa, Zhihe, Xuwei and Tianwu. I have been to Sweden for 4 years, at here, I got a lot of help from many friends, they will be my good friends forever, I want to thank them for all the time we are together, they are Mingtang and Chunlin, Jiangfei, Tangsi and their baby Tiantian, Xiaoqiang, Feifei, Liyuan, Haining, Luyu and Sijia, Zhangfan, Liushuang, Yanhong, Xiaowei, Yongxin, Tiegang, Guoqing, Mingchao and Zhouyuan. I would also like to thank all my relatives for being supportive in these years. They are uncles, aunties, cousins. I can always feel your care and support in every aspect. Last but not the least, my great gratitude to my grandma, and my father and mother for your love and belief in me that has endowed me with endless power to approach my goals fearlessly. And my wife Yadong, thank you very much for giving me so comfortable family life, and our future baby, I am very lucky that you will come just after I finish my PhD study. A good time for you. I love you forever!

Acta Universitatis UpsaliensisDigital Comprehensive Summaries of Uppsala Dissertationsfrom the Faculty of Science and Technology 1065

Editor: The Dean of the Faculty of Science and Technology

A doctoral dissertation from the Faculty of Science andTechnology, Uppsala University, is usually a summary of anumber of papers. A few copies of the complete dissertationare kept at major Swedish research libraries, while thesummary alone is distributed internationally throughthe series Digital Comprehensive Summaries of UppsalaDissertations from the Faculty of Science and Technology.

Distribution: publications.uu.seurn:nbn:se:uu:diva-205651

ACTAUNIVERSITATIS

UPSALIENSISUPPSALA

2013